From c1dc34f404369dd54476317344b751e676bb194f Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Sat, 30 Dec 2017 05:12:03 +0800 Subject: [PATCH 0001/1480] Update README.md --- README.md | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/README.md b/README.md index b866969b..aea54728 100644 --- a/README.md +++ b/README.md @@ -1,6 +1,5 @@ ### Deeplearning Algorithms tutorial - -最近以来一直在学习机器学习和算法,然后自己就在不断总结和写笔记,记录下自己的学习AI与算法历程。 +进来一直在学习机器学习和算法,然后自己就在不断总结和写笔记,记录下自己的学习AI与算法历程。 机器学习(Machine Learning, ML)是一门多领域交叉学科,涉及概率论、统计学、逼近论、凸分析、算法复杂度理论等多门学科。专门研究计算机怎样模拟或实现人类的学习行为,以获取新的知识或技能,重新组织已有的知识结构使之不断改善自身的性能。 * 机器学习是计算机科学的一个子领域,在人工智能领域,机器学习逐渐发展成模式识别和计算科学理论的研究。 From 44980317a7e590e50158ba1ac5b936ab7c220030 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Sat, 30 Dec 2017 05:12:32 +0800 Subject: [PATCH 0002/1480] Update README.md --- README.md | 2 ++ 1 file changed, 2 insertions(+) diff --git a/README.md b/README.md index aea54728..e45d1069 100644 --- a/README.md +++ b/README.md @@ -144,6 +144,8 @@ * [强化学习]() * [迁移学习]() + + #### 机器学习的基础 * 机器学习需要的理论基础:数学,线性代数,数理统计,概率论,高等数学、凸优化理论,形式逻辑等 From bf3a5a38edd0326a572124eb6e759c4a43a1b582 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Sat, 30 Dec 2017 05:14:14 +0800 Subject: [PATCH 0003/1480] Update README.md --- README.md | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/README.md b/README.md index e45d1069..306cb71d 100644 --- a/README.md +++ b/README.md @@ -144,8 +144,7 @@ * [强化学习]() * [迁移学习]() - - +算法模型的整体基本就是这样目录 #### 机器学习的基础 * 机器学习需要的理论基础:数学,线性代数,数理统计,概率论,高等数学、凸优化理论,形式逻辑等 From 31bf963a6134034da651c3be86385f6dd7caf9c7 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Sat, 30 Dec 2017 05:15:39 +0800 Subject: [PATCH 0004/1480] Update README.md --- README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/README.md b/README.md index 306cb71d..6896e9b8 100644 --- a/README.md +++ b/README.md @@ -144,7 +144,7 @@ * [强化学习]() * [迁移学习]() -算法模型的整体基本就是这样目录 +算法模型的整体基本就是这样目录,后续的算法模型我会不断完善和补充,更新! #### 机器学习的基础 * 机器学习需要的理论基础:数学,线性代数,数理统计,概率论,高等数学、凸优化理论,形式逻辑等 From b3d41d73ce9733aaee631578a913dca6f6e77a82 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Sun, 31 Dec 2017 00:13:57 +0800 Subject: [PATCH 0005/1480] Update README.md --- README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/README.md b/README.md index 6896e9b8..8f227627 100644 --- a/README.md +++ b/README.md @@ -1,5 +1,5 @@ ### Deeplearning Algorithms tutorial -进来一直在学习机器学习和算法,然后自己就在不断总结和写笔记,记录下自己的学习AI与算法历程。 +最近以来一直在学习机器学习和算法,然后自己就在不断总结和写笔记,记录下自己的学习AI与算法历程。 机器学习(Machine Learning, ML)是一门多领域交叉学科,涉及概率论、统计学、逼近论、凸分析、算法复杂度理论等多门学科。专门研究计算机怎样模拟或实现人类的学习行为,以获取新的知识或技能,重新组织已有的知识结构使之不断改善自身的性能。 * 机器学习是计算机科学的一个子领域,在人工智能领域,机器学习逐渐发展成模式识别和计算科学理论的研究。 From 6634b64ade9f93b3ba07f648744b6fc98a26502d Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Sun, 31 Dec 2017 00:15:19 +0800 Subject: [PATCH 0006/1480] Update README.md --- README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/README.md b/README.md index 8f227627..d116874a 100644 --- a/README.md +++ b/README.md @@ -63,7 +63,7 @@ * [关联规则]() * [支持向量机(SVM)]() -后面的算法我会持续更新整理算法和算法模型,后续的算法章节会不断的补上,希望可以对新入门学习AI开发和算法的开发者有所帮助! +后面的算法和我们的算法模型,我会持续更新整理,后续的算法章节会不断的补上,希望可以对新入门学习AI开发和算法的开发者有所帮助! #### 算法模型 * [回归算法]() From 7d407a7cb59b59fb943b055970fd067c38430875 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Mon, 1 Jan 2018 00:31:41 +0800 Subject: [PATCH 0007/1480] Update README.md --- README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/README.md b/README.md index d116874a..d2b76b9e 100644 --- a/README.md +++ b/README.md @@ -21,7 +21,7 @@ #### 深度学习 深度学习:深度学习是基于机器学习延伸出来的一个新的领域,由以人大脑结构为启发的神经网络算法为起源加之模型结构深度的增加发展,并伴随大数据和计算能力的提高而产生的一系列新的算法。 -深度学习的方向:被应用在图像处理与计算机视觉,自然语言处理以及语音识别等领域。 +深度学习的方向:被应用在图像处理与计算机视觉,自然语言处理以及语音识别等主要领域。 * [深度神经网络](https://github.com/AI-algorithms/Algorithms-Tutorial/blob/master/src/Neural%20Network/NNK.md) * [深度学习的入门]() From e1be9840f15eceeafdc308ef74881253e20d2eb2 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Mon, 1 Jan 2018 00:32:15 +0800 Subject: [PATCH 0008/1480] Update README.md --- README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/README.md b/README.md index d2b76b9e..d116874a 100644 --- a/README.md +++ b/README.md @@ -21,7 +21,7 @@ #### 深度学习 深度学习:深度学习是基于机器学习延伸出来的一个新的领域,由以人大脑结构为启发的神经网络算法为起源加之模型结构深度的增加发展,并伴随大数据和计算能力的提高而产生的一系列新的算法。 -深度学习的方向:被应用在图像处理与计算机视觉,自然语言处理以及语音识别等主要领域。 +深度学习的方向:被应用在图像处理与计算机视觉,自然语言处理以及语音识别等领域。 * [深度神经网络](https://github.com/AI-algorithms/Algorithms-Tutorial/blob/master/src/Neural%20Network/NNK.md) * [深度学习的入门]() From 89ba7f844696972ea52ed7d50983097ef42e0ee3 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Tue, 2 Jan 2018 09:18:48 +0800 Subject: [PATCH 0009/1480] Update ABT.md --- assets/src/ABT/ABT.md | 3 +++ 1 file changed, 3 insertions(+) diff --git a/assets/src/ABT/ABT.md b/assets/src/ABT/ABT.md index 2bd4d2de..c9e362d8 100644 --- a/assets/src/ABT/ABT.md +++ b/assets/src/ABT/ABT.md @@ -21,3 +21,6 @@ AdaBoost算法的特点是通过迭代每次学习一个基本分类器(即弱
,标记
,其中X是实例空间,Y是标记集合。AdaBoost算法的原理如下:
From a45b21755a93dc202bf3401e1aa822805ed4192d Mon Sep 17 00:00:00 2001
From: keke <2536485681li@gmail.com>
Date: Tue, 2 Jan 2018 09:19:12 +0800
Subject: [PATCH 0010/1480] Update ABT.md
---
assets/src/ABT/ABT.md | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/assets/src/ABT/ABT.md b/assets/src/ABT/ABT.md
index c9e362d8..a486632a 100644
--- a/assets/src/ABT/ABT.md
+++ b/assets/src/ABT/ABT.md
@@ -23,4 +23,4 @@ AdaBoost算法的特点是通过迭代每次学习一个基本分类器(即弱
#### 算法原理
-假设给定一个二分类的训练数据集T={(x1,y1),(x2,y2),…,(xn,yn)};其中,每个样本点由实例与标记组成。实例
,标记
,其中X是实例空间,Y是标记集合。AdaBoost算法的原理如下:
+假设给定一个二分类的训练数据集T={(x1,y1),(x2,y2),…,(xn,yn)};其中,每个样本点由实例与标记组成。实例
,标记
,其中X是实例空间,Y是标记集合。AdaBoost算法的原理如下:
From 85cf8aeede69d6945e49fde823ea4ccf5197177d Mon Sep 17 00:00:00 2001
From: keke <2536485681li@gmail.com>
Date: Tue, 2 Jan 2018 09:33:21 +0800
Subject: [PATCH 0011/1480] Update ABT.md
---
assets/src/ABT/ABT.md | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/assets/src/ABT/ABT.md b/assets/src/ABT/ABT.md
index a486632a..991b995e 100644
--- a/assets/src/ABT/ABT.md
+++ b/assets/src/ABT/ABT.md
@@ -23,4 +23,4 @@ AdaBoost算法的特点是通过迭代每次学习一个基本分类器(即弱
#### 算法原理
-假设给定一个二分类的训练数据集T={(x1,y1),(x2,y2),…,(xn,yn)};其中,每个样本点由实例与标记组成。实例
,标记
,其中X是实例空间,Y是标记集合。AdaBoost算法的原理如下:
+假设给定一个二分类的训练数据集T={(x1,y1),(x2,y2),…,(xn,yn)};其中,每个样本点由实例与标记组成。实例
,标记
,其中X是实例空间,Y是标记集合。AdaBoost算法的原理如下:
From 718e5ce33f8771ca8ffb6d1e532764b5ec7e7925 Mon Sep 17 00:00:00 2001
From: keke <2536485681li@gmail.com>
Date: Tue, 2 Jan 2018 09:34:06 +0800
Subject: [PATCH 0012/1480] Update ABT.md
---
assets/src/ABT/ABT.md | 2 ++
1 file changed, 2 insertions(+)
diff --git a/assets/src/ABT/ABT.md b/assets/src/ABT/ABT.md
index 991b995e..dfa76232 100644
--- a/assets/src/ABT/ABT.md
+++ b/assets/src/ABT/ABT.md
@@ -24,3 +24,5 @@ AdaBoost算法的特点是通过迭代每次学习一个基本分类器(即弱
#### 算法原理
假设给定一个二分类的训练数据集T={(x1,y1),(x2,y2),…,(xn,yn)};其中,每个样本点由实例与标记组成。实例
,标记
,其中X是实例空间,Y是标记集合。AdaBoost算法的原理如下:
+
+输入:训练数据集T={(x1,y1),(x2,y2),…,(xn,yn)};其中
,
From 9bf50a125da5c5f61ecab7c1b7850ba1100e28c1 Mon Sep 17 00:00:00 2001
From: keke <2536485681li@gmail.com>
Date: Tue, 2 Jan 2018 09:34:25 +0800
Subject: [PATCH 0013/1480] Update ABT.md
---
assets/src/ABT/ABT.md | 2 ++
1 file changed, 2 insertions(+)
diff --git a/assets/src/ABT/ABT.md b/assets/src/ABT/ABT.md
index dfa76232..eecb955f 100644
--- a/assets/src/ABT/ABT.md
+++ b/assets/src/ABT/ABT.md
@@ -26,3 +26,5 @@ AdaBoost算法的特点是通过迭代每次学习一个基本分类器(即弱
假设给定一个二分类的训练数据集T={(x1,y1),(x2,y2),…,(xn,yn)};其中,每个样本点由实例与标记组成。实例
,标记
,其中X是实例空间,Y是标记集合。AdaBoost算法的原理如下:
输入:训练数据集T={(x1,y1),(x2,y2),…,(xn,yn)};其中
,
+
+输出:最终的分类器G(x)
From 571bd7365ec8857cb8dfc7e92e4ad5702ce39376 Mon Sep 17 00:00:00 2001
From: keke <2536485681li@gmail.com>
Date: Tue, 2 Jan 2018 09:34:47 +0800
Subject: [PATCH 0014/1480] Update ABT.md
---
assets/src/ABT/ABT.md | 7 +++++++
1 file changed, 7 insertions(+)
diff --git a/assets/src/ABT/ABT.md b/assets/src/ABT/ABT.md
index eecb955f..6437a855 100644
--- a/assets/src/ABT/ABT.md
+++ b/assets/src/ABT/ABT.md
@@ -28,3 +28,10 @@ AdaBoost算法的特点是通过迭代每次学习一个基本分类器(即弱
输入:训练数据集T={(x1,y1),(x2,y2),…,(xn,yn)};其中
,
输出:最终的分类器G(x)
+
+(1)初始化训练数据集的权值分布
+
+
+
-
+(2)对于m=1,2,…,M
From 2815b1efbfcffeac7a6038fbb87c677a7edef420 Mon Sep 17 00:00:00 2001
From: keke <2536485681li@gmail.com>
Date: Tue, 2 Jan 2018 09:45:44 +0800
Subject: [PATCH 0017/1480] Update ABT.md
---
assets/src/ABT/ABT.md | 1 +
1 file changed, 1 insertion(+)
diff --git a/assets/src/ABT/ABT.md b/assets/src/ABT/ABT.md
index b000a409..ba692c95 100644
--- a/assets/src/ABT/ABT.md
+++ b/assets/src/ABT/ABT.md
@@ -36,3 +36,4 @@ AdaBoost算法的特点是通过迭代每次学习一个基本分类器(即弱
(2)对于m=1,2,…,M
+(a)使用具有权值分布Dm的训练数据集学习,得到基本分类器
From 2b96f66193eaf50ff4d14cf09d167d70c9652641 Mon Sep 17 00:00:00 2001
From: keke <2536485681li@gmail.com>
Date: Tue, 2 Jan 2018 09:46:03 +0800
Subject: [PATCH 0018/1480] Update ABT.md
---
assets/src/ABT/ABT.md | 3 +++
1 file changed, 3 insertions(+)
diff --git a/assets/src/ABT/ABT.md b/assets/src/ABT/ABT.md
index ba692c95..b8933966 100644
--- a/assets/src/ABT/ABT.md
+++ b/assets/src/ABT/ABT.md
@@ -37,3 +37,6 @@ AdaBoost算法的特点是通过迭代每次学习一个基本分类器(即弱
(2)对于m=1,2,…,M
(a)使用具有权值分布Dm的训练数据集学习,得到基本分类器
+
+
+
+
+
-
+
-
+
6AJq-_~Y-*r7C|bNM8+yIvZU5xOLG?1@V%UIDf7SI9S;G8r)KSW&n3Lxf zXlN$c^n-*%ai+828%mo#Aa-WDJri7#&9J|CNMt6?Gw1L)Q=a47*SrOu^aclW(%yxA zN@upKkIHE5M)TQ*XKBEY>+kmPOPd#WI>IsiJ5kxMrMwvdsE8j+_W1fSU;8IjFN ze*#BIL_t(oN9|W@cY-hsg*jy}m{Z(l_q2Oi;Q#-=rAe8sghvIBA4ZSI?I5|mxlL{$ z2*x!kFe)%AFoXhGv>XEFWEf()?Yp<^57C!G&-m wp8zE;l_XoIYGlFoz 9;0iLL?v)+p=ffA~a?xdZhx Q|9s&%iP+KxHIDs3QHZ z@UJ9$vte-jQA`$?Q>z9Qbx1+%Bmo-aw`4Ev3-3-KT3NDr;|zmhDnLLIhxCG(3eZG( za<&dsXLn%P>t^)A83rZ5@lJM-MQtfmdckM_f5c6)Ebe;_iD7WOQP2tSt4Q}pf4wA4 z% *>y6o)48x07+3kIsaN`xR7 p@2(~0va#t1$RB>~lG4Aq=NE1ZPMh!ivqAs>002ovPDHLkV1h}$KO6u6 delta 1300 zcmV+v1?&3f2DA%HiBL{Q4GJ0x0000DNk~Le0003|0000g2m=5B06jXG#sB~S0drDE zLIAGL9O(c60FzKmR7C&)000000024w0A>I>IsiJ50A^;9NH++OI*_)IklWjlff@nZ zk(nd}^Of0lk- MXc@!h{A8=n1|zD!xVgug-2`-lF2_T$aji?d9Cb_-QV8|}78 zI&jY#m5pPG0uc9^wyrOWqyja&>zGED$cbyJ^x0=^Z54fUsgi14m#3Gp2(49-tv6HY zUN=9_%^;PYZm46ab&ScQDl}NV6D!_A9mubi6IcPLc8g?r)WTUYX!s`XFO|-C#Bu^& zk1$1$=N9yT&Gr?wEaD^Pt-~XIbrX7-;hWg-7AnY0Y!dZgO0_Bek=p_SK94Z*_9gV# zHMB5&DLccy1f{ySx&x|5e68`RyYaCv*EU<5O8XL{i_C7pKX6+{z~>Q$q+Jf)h-Ofx zsHU#xuo8=;xu$$3osbdNrBcB`+0BA~;AU_SBtScV!x?$n>EVZR3@Nk04!Z^w$r83) zM31^0=2l}z&TQ9c$H@xMmg0vTG|}>QngFSkokmUO_!Kv(*a>TxB{-1Tt+{kZ0w#W; ze3xe~%W$|jyWK*-lfh8jNv;+8_jf !czxq3`izG{6+l zNi`dPe&QBNV1i$4T@eOfnwN5529=5m@X&QLU5DtXnUCo1(7x7BJ|c@;4ey?*!8AwM z)IFfyiU4V>vjpZj{K{KkQWyafKhS|NN-cpjjXRNo#E`O~Z-c$rVS5C`bF}Vyye{Fc zU&h+za^^uoSz6o!uvKsJqQA3Vphylg_;)6MVB&{9-5+NJf7DK@C9|+mN|ca&DV6ST z#jlz4SZ!Lo;p7u%6!SD<&BeFNzhz;x6SO rjFH^Q-gp {YA54Jb7h7Y0 zmkTW0?WRsIe>w9$OjF`%J;KEY@N#XXe1ugVHX8)j+RRBfO4ldL`1(=PfKy) lkH(V0utimP{`unaGQ?rclBW?-b% zKEuMonvD5w1Q>qZbud&5bd);#9dfP9@)=*LKkBZ3e5V++(T9eDvZ`mIT uD5G!JxUWF*C!<$(d{0|=?6Mc=7spyF7aq@pBTs1d;^AKG4 zUNLXyyC+|e9r`(5(!{Yx1-y;Z9^Z8qTbYfMuX8;MQC5*O8z;PuP)8xT@s5|1tI(Xl z;>P&W4%&CB|9F1`8Me!xt8W%gfbQWbz#W%4GV=&*{X*vO(2nyB<4xG8D_+#HJA$a_ zdgH=TNII|RtD>T-6%_}qaYRKY*DK17LKS_(uzK6LK~&(x75ESJaf)4MWL0AT0000< KMNUMnLSTX%Dr^@3 diff --git a/assets/images/82.jpg b/assets/images/82.jpg index ba27024f4ec22b854f7180598aef78e3bf6ec1ad..0f8fc925e2f8f7a9a4ae6e373005ec2347c422ef 100644 GIT binary patch delta 1177 zcmV;K1ZMl^28RlNiBL{Q4GJ0x0000DNk~Le0002!0000g2m=5B0FGDZ^#A|>0drDE zLIAGL9O(c60HRP#R7C&)000000024w0A>I>IsiJ50A^+YX0|#206G9VIsj%mI%YaL zkUD0zI*^b$klSVe0A>I>W&mbpIsj%mkY;9PW{}%vwzg&(w*Qbi0FXL5kUD0NW&n_8 zI*^b$khYMJ+uM #m1k+l?jb>3EU@LLAr81CP=Jb+Qp{adHzqc>VF& zl~85$I7M?Ib8#J2KvgnY5Xic$)7{(Km?EmHgq0%v4nVJO^!D>bOgM#TG9VsjHCcZt zS_q~>z)>wo-rNDCt5Qe+*`7yLqKHr?Pp%xKGPX$e*xDjVqvD;(g7eEXs>) ^DdUfVYn-!ckMa1rCCB%| zu(bWQdl=jVfb8-T3d)p?g1&7EAEOT?8>DC()QNeLl+dd3I!Sj$S2dxmg0qschO&~V z)cl317${{c-4?ics6(62Dy7qB4IeqVo@ZWNp-to5>UaT9Lw>-)N`Wd}ktcs`nz_fd z#YQwQ-sEseQjS9W* (o##)g3MfPyn@o}!Pw68nF&1$xjky9de>!%8<9ilLWR^T3J(s87D9|EkiP`Slh; z(ifL}Z>*(=QfS8Q!Gxplf|~~Ky`Y9D#oONfsw;{c9%mE5d{B4Uhjo{mdX4lFVj?&U za}lm}-~ LJ@s*_J8vzVljZvn*M z2hO@=?6<#J4tfHm<$?*cumINXD@UgaRfn5@3w_kSYIN1O7W>k)&+DVsECpQt@)rKg zSU$mAlM;hY%Ord!Bm93re;Ian0GS%sVo#d(xs%bL3knC+sr=70(xuKa0^IA|o&WqX zdJ2QHJB>`vx{#|jA6lk{S 32vn|ERWIuQaKK!(d#yg>P&5`IB-xLLY~ rW&FP{;EU5eF1CA|bKH9LfbW6-mg;Z+RBMp700000NkvXXu0mjfN609C delta 780 zcmV+n1M~cc3g-rYiBL{Q4GJ0x0000DNk~Le000340000L2m=5B0Co3!%K!iX0drDE zLIAGL9O(c60EbXaR7C&)000000024w0A>I>IsiJ50A{v20024wIywMmIy#U#X0|$z zkUEgtW&i+Y06Jy>W@b8&W{}%vw*Qbi0FXL5kUD0NW;&1>wvdq9|F&iTwq`oEW@fgK zkhZpFwzmJa|F)5|Bm|$?AQ_R|Nq+)INkl XTw zzNJZ-t%OGfj~_;l$L%1wy}3 <`hGLeKc@ zVvPtg)a!r)%`P`RDV#oBO)!TdooOLrUqVId9pbyi@$WZD {V$Z-bE ?8pi W^0Dr_y zvMla<4vArKyiw2z@T*AoM}NH}P0Z}z3f%mpCa=? cqPM1k=cVhr<# gmygF3!JmE*BBzy3-kB4gFu{oawq|bFG$sg<9rVKfLVms ze;SC=E4u9L#2XC=zzYVdy-I{281Jqn=CZNr2*@9P?UK^Jx#t&d3r?Hw{Ify;0000< KMNUMnLSTZBTVGxP diff --git a/assets/images/83.jpg b/assets/images/83.jpg index 0f8fc925e2f8f7a9a4ae6e373005ec2347c422ef..680417893c7f5e26ff8ee5d54f49503bc3b20a4b 100644 GIT binary patch delta 944 zcmV;h15f;i3V;ZIiBL{Q4GJ0x0000DNk~Le0001{0000g2m=5B01hP3<^TWy0drDE zLIAGL9O(c60FzKmR7C&)000000024w0A>I>IsiJ50A{v20024wIywMmIyyQ!I%YaL zkUD0NI%c*yklSVe0A>I>W&mbpIsj%mkY;9PW{|dKklSV>w*Qbi0FXL5kUD0NW;&3N zI*{AjklX*ZW&pNkI<{tJwvdpKt{M^l|F&lT+mQeNw*UXz*3;W8k=#jt0@O)FK~z{r z?N@7elrRtlYZt8#uvUd#ZBtt}K9(f^|JONrK{5#q?7F}?Hm6^j4dl+9J9maGdtT2B z9M3?TUmh~i Wn@*H3b z;8E}w20htbZNgky-<6J{Q#CXso&mOarOt^4S#z*~ft)N)(Et;F{DYshA+}m=fnk#U zutK(Q)11sFiR5dDEdsR5<0i+v0@Oqyw^;``W}&hId5)Uf;9OoCHk?>o4R+gxEeD2S zi|DKLe`5EeDNR#J6%hn7s@i$A$=I8ZLA$3h@oR2e7-q+4vP&_w{W==0mswVf88xK0 z@hUJl(EvZtCVa4e5g5vHEN9*CU4`3U29qwDl)1{FREb7H=9P>QYas$dxw%EvE749B zDM!kpSqlnDMH}fBx>#_X+W$pkVQ99DG%c%Qd)t)kORv`9`b~>MJu8oqPnUmg@T{W6 zfQry? 0drDE zLIAGL9O(c60HRP#R7C&)000000024w0A>I>IsiJ50A^+YX0|#206G9VIsj%mI%YaL zkUD0zI*^b$klSVe0A>I>W&mbpIsj%mkY;9PW{}%vwzg(nw*Qbi0FXL5kUD0NW&n_8 zI*^b$khYMJ+uM-a|F&iTwq`oEW@fgKkhZpFwzmJa|F*XO+qVDz+mJfjwr1P5w%gl~ z+yA!P|J&RD|NpjT|J#uN|F-}C+eUECy^+mHe*?@(L_t(&L+x6La?~&kb!fweqa+k4 z$(F08Ts5#XiU0rCkSr&$9Ou~1bY| TJg=jJ$9%nUK ze<@lBrb56`ElA$n0i>%^NC4TMM^&PTP$o~V9HcU~NcPy;B1xm-oymgp%QULWWMH67 zqs> }Iofa5n0 z_rkEW{kD4;+ysE^@)8Qll#PPEZ3-Ww4<#F nlxOu2Uo6jnx(`OAIIk}!^UR|M0 qG%w!da7j{*Lj2iZB&{!hRk=H{^|_1tLvnK|)MWEH1}KbU>(Mq-BIazFe!yxm z4#$XfKLceriG==W3*!fjSaW8~GM pIzmr|33oDey9iAB0?w4Lic(*W z{0`Pxf;Yy7fbxKXGi;uskG>N7f3yX9&@;OS$`iv%HyMhdmsj(^iUg=nzNi1H(wq79 z7DCb&mwa!mrHE2!#_hp`qwj*72JgL~hA73`-u %5+P}!4jON5*y`{Hbs@!;l)=;BVe-h}0Cflea z{k<^b`3D3N*tmiMG3X>YWKpCaxCNQptvZEdwyYjtU1De33r(bq8my|5PbIUMq>*m{ z#NY?cx@7FPzgZ4?0;T1G3AC^P*6u4urwUbvn|}*^)V^wT)wmY>(zMU(qt+}1T>kPF z{>)fD!CaFPgHFpNd?zFPe?Wg3c6R`o8rNb^n)bPq(Vz )f6H z{4sh8gR?u0OwPKHt2Q56riNL}q-h_Ub^FtSB#^39ayAKcv9tJ42A#V#uk)9Qj{U2E zXR#}7_IbCLnMCy;AxS=lwe0XXyx@11J+6-r%K?68!Rg!%AQs3hX>UFlu(fxihhczh z? *S>dz^FJdh~$rf&Z53Z~s(lkhTB-002ovPDHLkV1nGVQ-=Tm diff --git a/assets/images/85.jpg b/assets/images/85.jpg index 680417893c7f5e26ff8ee5d54f49503bc3b20a4b..b99919143b6b25f330656c1e85d5e1299a916608 100644 GIT binary patch delta 1381 zcmV-r1)BPR2!;$YiBL{Q4GJ0x0000DNk~Le0003u0000g2m=5B0O7;z@c;k-0drDE zLIAGL9O(c60GyFBGaisSklSVe0A>I>W&mbpIsj%mkY;9PW{|dKklSXqk$pcD+uQ&D z|F&lT+mQeNw*UXzfiJ#1k 6bADv=<;6BB zu2yqC!!dNPKHff5uk$nA2=LjaSa=K2DmUNm?|xoyr?6R(P9gQ6GFgg$LG03mM+CL+ z+u~|dLX4YY%^o-oOtoHEVXRtjlKSSUTqS_&B&)f7m_+WOgx2r4{@|%5Kn+AShMcR` z77<6(YAPa+sA89>Lh_6)E=&Tj$+H2PX1=S=r1o`&)!HFfap32TdbYB4LpQ{Y`k0Wg z>iLW#9MkClLfk<;r^)?)l!N(5EjA_}M;2UXhVINRMkXhm@2K{beF&*=0x}>oLW1|P zi)n^rvJgiYTg3=obtBIz= 2uLbw(K zV>87fp?8x*h~?Fg=eQwHwv{4|u5gm)rUdueg+#WntFwJWoW{U^I3CF&K|`i9lb0w4 zrrZ`MNeln@LU6xbDC#(hx&79zjMEtCoivMha5a2bn!nD#2^e<-MF`MKE|n`7d|o~= zEY=~s;D<-(4dG9W*9s@R39Uj3mdD>wJQ~}{S@7Vfu@UOJv7H?CHL_s~oIHeqhB6k% z&0xGF!QHPsm1Q@7=_iN;G=b=cO}@e(=-b%a>n)ej`k$;8NWqk&c8d?-mXuLBwa!*a z_42)2M>t+DpcA?^1{z9R#0^i4-7((OlD|+rU(gUJRQ>})0FPm4e;y<_Qc94bIw_d4 zOSkwCZq?caBjl^7W}Uh9m5G7g6;N8nF~m(2LV~&Noh=Q2)VX{|686O&SMp7Gr)b`k zAhn$oEb-Y hxj3+j9cpLk${CDFF0t^iF2C18Yh8f>* zWV*Qg%DXFpJ7&^C%oxZzK_yBCSxh+D<(A)E@~KXnicnscV8Sjv;t@Biev@|h_~l&y z!5coAtJW(x<8PTv6b9B+aR>tq9e&*O>zco}F6AtLeFYHP2+L2}3kkt7^; R+ z7<+<6xKWH{T_sr*P i@VRVd}w67G_ zXygF8M7hZU^ceJrHlRk9AcGzf-ZZ6~j;JOfO1{r>OOWXgHShOXVq$1c!34g>dj|{f znA9C-?3uqm<-IY~*>)m_%GEqQ3@l(>Jk;56GiQ_?hf(0JxxcRmYNa^HaTi`RdjWZL zcX&G(x2e8AF<7D9H?q&|I8~#UCc$1{EZk=QwVDo?b=9x@)PFOyyOmCVN2nJVi*F~4 nUZ@wq2}^pyXFeb8v=R6ZINhldfHCTz00000NkvXXu0mjf{RWW- delta 859 zcmV-h1El 4 p?};STnBd&k~2 z>mifo!2)n?eSf=qyxyM>>A!sx1bE@1(SrQGr&o0g`doj_VnyyrkQH^tZw~DyBjsv9 z0uV^Mi${PWiDZ-&9?gqGCi#my2LxG9;ds>WO1k9w7nbK*5raIJ%CxLlyGSaj#CN3` z=mch@!B7twd*v2Mh~j+m9AFFJQScWAJ=tAt!dzP4m5!oQH8doi0k(Lh&WQ$DbFhJd zoGee#026=wgP*k_wpwk0VUqo@Lbh H7YX?&%q;Xz^kb@z^^gzpBK2Cpo8e30uGtP_9JNWfpY^Cq~5nsI^D~h80kgJfa`9>UX+@p7J_e!y~wtR08Yx(&WgW912@>VdmoGwut2=L zdWpUW&I${?T*L5Ii93vW1drk~!t4j89%-4yJIsPG2N>yFNtg8SGjK*W2d;{CnANcv l?J)8?Ys`GVJjwqPg1^+FPFE#WI Dz_y~!Ns7YXyJ zIh2=CPEqM7Mwr7&%OO&t %8N06>650!R=5!iYdB5&)8Z0m%>v zL;@y35D5~2zpWqug1;OPhya2?qz@w@g-8JO=K?bOWfwsrDGUNINP !3Ks z%G;2pCd 9~VncyP$+1*>QPK)-aN+WCapFC5K`DYjL0 zh~K48Pb1RR;)9MCIo2CVrI91|x(3f}k_qpd0m<;*rulm*ex5npSDLAKnNW208C!v` zjlNsk@b=duv8PUamh@N~@ejw^+&5>{+DzE^)PBZ~%Y7WF^i5Rk+fA7Kuy~3YSyjil zLf2g>wP0MZo9|Djp&J)PZz8-Y?Bbwj+C*j7My7Az&7+YyYwxO$>7SX~RU=3#pnfPz zA6jhwDXuoKRiavDaBGGzYBXWQZoSLApnM)>iEV1;DpS{c8&w3h)jYrPB`#xdRQE`< zc>2;wxFOD-(0ysG0IHk@@MoU>2Z^+3m0D{gIpY&C%4_U oLhf6W_Yn-aVglwuVCRqc zvTj@jhw9Rb)uc3I)&=8x^vV*h+MkyULetHvJ*V7-N%}Hs&Tnx_8Byv>Bz>?(A0zxV!SVC<=KGCq2nqMZJuVVqU%4TuXh)_7$s! zKt(P&JvGAu^YB@~-CdpIwPa2RVHS5DZJ9cg>DJb@V@5PAdT*Nla_!npjALhRdHh8c zabB+HUyaWXJFP6;^P>@&drCL8Sjld6_Ul%1zcg+>A x7=vUG-=n?yjn# hxa zyGxM_s9SeQ&-jOoD&OdHcp{n7qS_Q*KZK_$7|i$YoH)W} E!f(J7{pv!Fk8#O}B zsjKl{S0(;H=Qf+dnjvH2RAiy6hV0OxXt{MtTBpXA-xswvnMq7i)~kx-p16-H%6FQt z_}P-lLNk4e -`8ug&wb18%Xu{Nw>;UADa-oAX9}|7KsmXb+Y?cRxqC#^@+i6-Z7PK8yd7w16l0i zJGB}^VQ+yJs-PFOj9or&U3xH2Crw*@!m^?`n`fZw-1Z{I1)5c{TKvig3JRllqT3!M zZBYDJAk+$-FKj;`xo0jMiPy_~JV@TZ(Imj1EF8gy=4xeniur=}+6@sE#7EIvn3hSX ze^?2L`>;*bZ3oPK%f<-}FBXW0%JU66nE@taptP|_xgc(^_Pg*j{!S~0ewxPHg1s1t pV>5YMI^ Date: Wed, 3 Jan 2018 09:04:48 +0800 Subject: [PATCH 0022/1480] Update ABT.md --- assets/src/ABT/ABT.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/assets/src/ABT/ABT.md b/assets/src/ABT/ABT.md index 56518389..13ebc1de 100644 --- a/assets/src/ABT/ABT.md +++ b/assets/src/ABT/ABT.md @@ -43,5 +43,5 @@ AdaBoost算法的特点是通过迭代每次学习一个基本分类器(即弱 (b)计算Gm(x)在训练数据集上的分类误差率 -
From 3641780172a6616ff44f00c5688b6831a23f9177 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Wed, 3 Jan 2018 09:05:24 +0800 Subject: [PATCH 0023/1480] Update ABT.md --- assets/src/ABT/ABT.md | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/assets/src/ABT/ABT.md b/assets/src/ABT/ABT.md index 13ebc1de..ba7347bb 100644 --- a/assets/src/ABT/ABT.md +++ b/assets/src/ABT/ABT.md @@ -45,3 +45,8 @@ AdaBoost算法的特点是通过迭代每次学习一个基本分类器(即弱+
![]()
+ +(c)计算Gm(x)的系数 +
![]()
+
From ccf1e4890476541c7af1488c49d488c51a9dd52c Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Wed, 3 Jan 2018 09:07:39 +0800 Subject: [PATCH 0024/1480] change page --- assets/images/84.jpg | Bin 1024 -> 727 bytes 1 file changed, 0 insertions(+), 0 deletions(-) diff --git a/assets/images/84.jpg b/assets/images/84.jpg index 680417893c7f5e26ff8ee5d54f49503bc3b20a4b..de1f93b5bcea399ad289fffc3493e85ece90fe72 100644 GIT binary patch delta 638 zcmV-^0)hR22-gLFiBL{Q4GJ0x0000DNk~Le0001d0000g2m=5B07e(jW&i*H0drDE zLIAGL9O(c60BKN6R7C&)000000024w0A>I>kN{@3IsgDV06IE4W;!~MI%c*yklSVe z0A>JYW;y_7I*?{&W@fhkkU9X6Iy#U#W{}(ewq^jfkdU^KsUQTxoct1z%}IX&)k#D_ zR9HvtR)0^zFc58Zn+j88%A_jL?fZX7-?hWA4%hZBfy8iO34h$(yPvP8)J<*`n5{s& zj20+CCF+
CJJKLf z&mPXfuS(h rtoGk28h _B|kh1f}?n{ZA@@}F+EJ;w-tTL_Qk&@r8$$s#?L!{N#z6X~%WyeBz%``0eZVy>nK zc(RM+s*8{1QzXYGxPVeccr1@#hG#Av9Hj6Mj^)%+pVf= DiBL{Q4GJ0x0000DNk~Le0001{0000g2m=5B01hP3<^TWy0drDE zLIAGL9O(c60FzKmR7C&)000000024w0A>I>IsiJ50A{v20024wIywMmIyyQ!I%YaL zkUD0NI%c*yklSVe0A>I>W&mbpIsj%mkY;9PW{|dKklSV{w*Qbi0FXL5kUD0NW;&3N zI*{AjklX*ZW&pNkI<{tJwvdpv|F*XO+mU`D1lH5rERmy0e*)A=L_t(oN9|W@ca$&? z1#1_r53p8+U2RiaH$Ijm|NqxHc|kG>4eYwWIX0(XnhoU6ojZ4iEPGzh3>?oun_nI> z(d6~ HhAjt%VT 5o;j=L%F#{)hp3X6)8u`qFD `SlK;rdOBLOm;wkx!R@Zt$$4#ej;?aO1#G zJDU2=H)V}eJ^iG7Gq_ F-mU3{zkZNps;L-XC2GtO};=P}oN z+BzS87bOHIc7D6_wg-IE5zVxo9{OO~-YGcK+XMXhN4@PRTa{}EQzoQwUWAZ?A;a`Q z%U?cDe|s8RPjfTQi_kmx^ 03#c^zbuqMm7hoiguXQu^H_!@;htHe7`)&{}Y10)S^yTB~?1G00000 LNkvXXu0mjf1Zb%> From daac494fbf35e60f977ffda6c3f2aee4f9056380 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Wed, 3 Jan 2018 09:08:51 +0800 Subject: [PATCH 0025/1480] Update ABT.md --- assets/src/ABT/ABT.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/assets/src/ABT/ABT.md b/assets/src/ABT/ABT.md index ba7347bb..0bf95596 100644 --- a/assets/src/ABT/ABT.md +++ b/assets/src/ABT/ABT.md @@ -48,5 +48,5 @@ AdaBoost算法的特点是通过迭代每次学习一个基本分类器(即弱 (c)计算Gm(x)的系数 -
From 8844133c971dad4027f3e0834028447afb683de8 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Wed, 3 Jan 2018 09:09:08 +0800 Subject: [PATCH 0026/1480] Update ABT.md --- assets/src/ABT/ABT.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/assets/src/ABT/ABT.md b/assets/src/ABT/ABT.md index 0bf95596..d87bfa34 100644 --- a/assets/src/ABT/ABT.md +++ b/assets/src/ABT/ABT.md @@ -48,5 +48,5 @@ AdaBoost算法的特点是通过迭代每次学习一个基本分类器(即弱 (c)计算Gm(x)的系数+
![]()
-
From bbd0213978c4cf1de6993c21cec444f91b2efcb4 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Wed, 3 Jan 2018 09:12:21 +0800 Subject: [PATCH 0027/1480] change page --- assets/images/86.jpg | Bin 0 -> 1378 bytes assets/src/ABT/ABT.md | 8 ++++++++ 2 files changed, 8 insertions(+) create mode 100644 assets/images/86.jpg diff --git a/assets/images/86.jpg b/assets/images/86.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d83e398b5e7bf094183494c9fee3f0a4708f35b3 GIT binary patch literal 1378 zcmV-o1)chdP)+
![]()
Px#nNUnrMF0Q*0000006G8wW&k=m06LHWW@Z3pwmJX+IsiI40A@NmW;!~MI%c*y zkdQi%+hza&W&k>70A^-70A@OnW@Z3pW@cuP+h(@gX14#3IslM5I*>YMkdQi%+uM-a z|F&iTwq`oEkdU^vX12Efw*R)a|J%0z|J#r{+qP!gwzk{bklX*Z+yC3!|NsBCX8+re z|NplC|J%ZVKCl1)00DGTPE!Ct=GbNc000SaNLh0L01m_e01m_fl`9S#0000PbVXQn zQ*UN;cVTj608MFQa&L2QW^^D=W@c$)Wq<88aR2}WFiAu~RA@u(S&4GfFc7tyhJ>TF z2~e h7$r1~R>fQ%y2Q~5#V$)ZFaluopfn$)^U{ R?ct;V;71 zQ7|OtYF)Jg!&R@XTOo(vHcSJmQocgpFk=JVo>4raV{pFe;b$5_!h>7NZZ|st%yF;* z`;_AKyC}#Lw9sM}-!z8|M7gg1lpohD?n-GkXj$DOk&=!9{@hM<0$Kxbjtox|kBWGX zS6o|O2ojotfzlIR7=vaJ0IMH4%f~Km7z?^p$Uqi!8gsd>gpJ{ =y;ASM z3n$(Jw4>Y($8G5* 3~8${*Xt?=iGUI>bYEBC z1O7mgq|^b%GIV=D V zN{dGhavh%CQj#usP>XMU@7C5&vRuJY+WVT`*_D9}tx__C_-C=H#iIrdNd!$@r<5!q zX?L}eNBd^)Yk6l21u~r7mqJs@*rV@c>WPL5nc9`v?C+9?vvGbRM;jVvA}{8(EfoR( z&*D*+Mn&LRs^=&-W@Ri@2&o7CKfGC;U|MHQp);)~bK2b$I&<=SQeQ2hGpQ%@*}Nfi k=Hs~c)Fa>#_+JqC2V`D>wcK;qPXGV_07*qoM6N<$f~g98djJ3c literal 0 HcmV?d00001 diff --git a/assets/src/ABT/ABT.md b/assets/src/ABT/ABT.md index d87bfa34..962cac88 100644 --- a/assets/src/ABT/ABT.md +++ b/assets/src/ABT/ABT.md @@ -50,3 +50,11 @@ AdaBoost算法的特点是通过迭代每次学习一个基本分类器(即弱 + +(d)更新训练数据集的权值分布 +
![]()
+
++
+
\ No newline at end of file From 89c7a532d6982c78d829de95671b549556b5aa0a Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Wed, 3 Jan 2018 09:13:00 +0800 Subject: [PATCH 0028/1480] Update ABT.md --- assets/src/ABT/ABT.md | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/assets/src/ABT/ABT.md b/assets/src/ABT/ABT.md index 962cac88..fd92f3c2 100644 --- a/assets/src/ABT/ABT.md +++ b/assets/src/ABT/ABT.md @@ -53,8 +53,8 @@ AdaBoost算法的特点是通过迭代每次学习一个基本分类器(即弱 (d)更新训练数据集的权值分布+
-
+
![]()
-
\ No newline at end of file +-
+ From 8647d4d045f916fa74fae7b09437ca05e57ea523 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Wed, 3 Jan 2018 09:13:36 +0800 Subject: [PATCH 0029/1480] Update ABT.md --- assets/src/ABT/ABT.md | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/assets/src/ABT/ABT.md b/assets/src/ABT/ABT.md index fd92f3c2..df2d287c 100644 --- a/assets/src/ABT/ABT.md +++ b/assets/src/ABT/ABT.md @@ -58,3 +58,8 @@ AdaBoost算法的特点是通过迭代每次学习一个基本分类器(即弱
+ +其中Zm是规范化因子 +
![]()
+
From 9394dbbc7f857b40fe718287918bcf534f4bc19b Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Wed, 3 Jan 2018 09:13:52 +0800 Subject: [PATCH 0030/1480] Update ABT.md --- assets/src/ABT/ABT.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/assets/src/ABT/ABT.md b/assets/src/ABT/ABT.md index df2d287c..95f87105 100644 --- a/assets/src/ABT/ABT.md +++ b/assets/src/ABT/ABT.md @@ -61,5 +61,5 @@ AdaBoost算法的特点是通过迭代每次学习一个基本分类器(即弱 其中Zm是规范化因子+
-
From 5579bd99565299117c45c53ab98dbf11e7cc810d Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Wed, 3 Jan 2018 09:14:46 +0800 Subject: [PATCH 0031/1480] Update ABT.md --- assets/src/ABT/ABT.md | 10 ++++++++++ 1 file changed, 10 insertions(+) diff --git a/assets/src/ABT/ABT.md b/assets/src/ABT/ABT.md index 95f87105..a8f2852e 100644 --- a/assets/src/ABT/ABT.md +++ b/assets/src/ABT/ABT.md @@ -63,3 +63,13 @@ AdaBoost算法的特点是通过迭代每次学习一个基本分类器(即弱+
![]()
+ +(3)构建基本分类器的线性组合 +
![]()
+
+ +得到最终分类器为 ++
+
From ed3cce0f3bbdcd6fc953333694fc162054eb424a Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Wed, 3 Jan 2018 09:16:08 +0800 Subject: [PATCH 0032/1480] Update ABT.md --- assets/src/ABT/ABT.md | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/assets/src/ABT/ABT.md b/assets/src/ABT/ABT.md index a8f2852e..d7d47252 100644 --- a/assets/src/ABT/ABT.md +++ b/assets/src/ABT/ABT.md @@ -73,3 +73,10 @@ AdaBoost算法的特点是通过迭代每次学习一个基本分类器(即弱+
+ +#### 相关应用 +AdaBoost算法主要用于分类问题。它是通过改变训练样本的权重,学习多个分类器,并将这些分类器线性组合,以提高分类的性能。它的应用范围非常广泛,可用于二分类或多分类的应用场景,可用于特征选择等。 + +#### 优点 + +优点:分类精度很高的分类器;弱分类器容易构造;算法简单,不用做特征筛选;不用担心过拟合 From e3e7044431a354e5ae95d57612392f204eaed5d0 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Wed, 3 Jan 2018 09:18:04 +0800 Subject: [PATCH 0033/1480] change page of CBA --- assets/src/CBA/CBA.md | 6 ++++++ 1 file changed, 6 insertions(+) create mode 100644 assets/src/CBA/CBA.md diff --git a/assets/src/CBA/CBA.md b/assets/src/CBA/CBA.md new file mode 100644 index 00000000..39a1d72e --- /dev/null +++ b/assets/src/CBA/CBA.md @@ -0,0 +1,6 @@ +### Deeplearning Algorithms tutorial +谷歌的人工智能位于全球前列,在图像识别、语音识别、无人驾驶等技术上都已经落地。而百度实质意义上扛起了国内的人工智能的大旗,覆盖无人驾驶、智能助手、图像识别等许多层面。苹果业已开始全面拥抱机器学习,新产品进军家庭智能音箱并打造工作站级别Mac。另外,腾讯的深度学习平台Mariana已支持了微信语音识别的语音输入法、语音开放平台、长按语音消息转文本等产品,在微信图像识别中开始应用。全球前十大科技公司全部发力人工智能理论研究和应用的实现,虽然入门艰难,但是一旦入门,高手也就在你的不远处! +AI的开发离不开算法那我们就接下来开始学习算法吧! + + +#### CBA \ No newline at end of file From d9a36e22fd1a2539367aed8b1e498afc109daed6 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Wed, 3 Jan 2018 09:18:47 +0800 Subject: [PATCH 0034/1480] Update README.md --- README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/README.md b/README.md index d116874a..8a458a5d 100644 --- a/README.md +++ b/README.md @@ -55,7 +55,7 @@ * [Prefixspan](https://github.com/KeKe-Li/tutorial/blob/master/assets/src/PN/PN.md) * [PageRank](https://github.com/KeKe-Li/tutorial/blob/master/assets/src/PRK/PRK.md) * [AdaBoost](https://github.com/KeKe-Li/tutorial/blob/master/assets/src/ABT/ABT.md) - * [CBA]() + * [CBA](https://github.com/KeKe-Li/tutorial/blob/master/assets/src/CBA/CBA.md) * [KNN]() * [Hopfield神经网络]() * [决策树]() From ab9334f15c1b7b87ecc3ae7fdcc8265cd30a7a62 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Thu, 4 Jan 2018 09:17:50 +0800 Subject: [PATCH 0035/1480] Update CBA.md --- assets/src/CBA/CBA.md | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/assets/src/CBA/CBA.md b/assets/src/CBA/CBA.md index 39a1d72e..19127937 100644 --- a/assets/src/CBA/CBA.md +++ b/assets/src/CBA/CBA.md @@ -3,4 +3,5 @@ AI的开发离不开算法那我们就接下来开始学习算法吧! -#### CBA \ No newline at end of file +#### CBA +CBA(Classification base of Association)算法是一个基于关联规则进行分类的算法,该算法首先利用Apriori算法挖掘出的关联规则,然后进行分类判断。 From 6b52bda45e125286bff26339beeb40cb04ffa7e4 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Thu, 4 Jan 2018 09:18:11 +0800 Subject: [PATCH 0036/1480] Update CBA.md --- assets/src/CBA/CBA.md | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/assets/src/CBA/CBA.md b/assets/src/CBA/CBA.md index 19127937..bba4ed82 100644 --- a/assets/src/CBA/CBA.md +++ b/assets/src/CBA/CBA.md @@ -1,7 +1,13 @@ ### Deeplearning Algorithms tutorial 谷歌的人工智能位于全球前列,在图像识别、语音识别、无人驾驶等技术上都已经落地。而百度实质意义上扛起了国内的人工智能的大旗,覆盖无人驾驶、智能助手、图像识别等许多层面。苹果业已开始全面拥抱机器学习,新产品进军家庭智能音箱并打造工作站级别Mac。另外,腾讯的深度学习平台Mariana已支持了微信语音识别的语音输入法、语音开放平台、长按语音消息转文本等产品,在微信图像识别中开始应用。全球前十大科技公司全部发力人工智能理论研究和应用的实现,虽然入门艰难,但是一旦入门,高手也就在你的不远处! + AI的开发离不开算法那我们就接下来开始学习算法吧! +机器学习是一门多领域交叉学科,涉及概率论、统计学、逼近论、凸分析、算法复杂度理论等多门学科。主要研究计算机怎样模拟或实现人类的学习行为,以获取新的知识和技能,重新组织已有的知识结构,不断的改善自身的性能。 + +机器学习理论主要是设计和分析一些让计算机可以自动“学习”的算法。这些算法是一类能从数据中自动分析获得规律,并利用规律对未知数据进行预测的算法。简而言之,机器学习主要以数据为基础,通过大数据本身,运用计算机自我学习来寻找数据本身的规律,而这是机器学习与统计分析的基本区别。 + +机器学习主要有三种方式:监督学习,无监督学习与半监督学习。 #### CBA CBA(Classification base of Association)算法是一个基于关联规则进行分类的算法,该算法首先利用Apriori算法挖掘出的关联规则,然后进行分类判断。 From 8548fedd70311c7191613152917d681b054ee168 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Thu, 4 Jan 2018 09:18:30 +0800 Subject: [PATCH 0037/1480] Update CBA.md --- assets/src/CBA/CBA.md | 2 ++ 1 file changed, 2 insertions(+) diff --git a/assets/src/CBA/CBA.md b/assets/src/CBA/CBA.md index bba4ed82..bc1328e7 100644 --- a/assets/src/CBA/CBA.md +++ b/assets/src/CBA/CBA.md @@ -11,3 +11,5 @@ AI的开发离不开算法那我们就接下来开始学习算法吧! #### CBA CBA(Classification base of Association)算法是一个基于关联规则进行分类的算法,该算法首先利用Apriori算法挖掘出的关联规则,然后进行分类判断。 + +CBA算法作为分类算法,它的判断依据是Apriori算法挖掘出的频繁项。如果一个项集中包含预先知道的属性,同时也包含分类属性值,然后计算该频繁项,能否计算出由已知属性推出决策属性的关联规则,如果满足规则的最小置信度的要求,那么可以把频繁项集中的决策属性值作为最后的分类结果。具体的算法细节如下: From 09ee75578b737e1d5db95ec9df10320985853676 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Thu, 4 Jan 2018 09:19:01 +0800 Subject: [PATCH 0038/1480] Update CBA.md --- assets/src/CBA/CBA.md | 2 ++ 1 file changed, 2 insertions(+) diff --git a/assets/src/CBA/CBA.md b/assets/src/CBA/CBA.md index bc1328e7..62bd55b9 100644 --- a/assets/src/CBA/CBA.md +++ b/assets/src/CBA/CBA.md @@ -13,3 +13,5 @@ AI的开发离不开算法那我们就接下来开始学习算法吧! CBA(Classification base of Association)算法是一个基于关联规则进行分类的算法,该算法首先利用Apriori算法挖掘出的关联规则,然后进行分类判断。 CBA算法作为分类算法,它的判断依据是Apriori算法挖掘出的频繁项。如果一个项集中包含预先知道的属性,同时也包含分类属性值,然后计算该频繁项,能否计算出由已知属性推出决策属性的关联规则,如果满足规则的最小置信度的要求,那么可以把频繁项集中的决策属性值作为最后的分类结果。具体的算法细节如下: + +1.输入数据记录; From 111893244589b71ad04e1578b11d429575f1a02d Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Thu, 4 Jan 2018 09:19:19 +0800 Subject: [PATCH 0039/1480] Update CBA.md --- assets/src/CBA/CBA.md | 1 + 1 file changed, 1 insertion(+) diff --git a/assets/src/CBA/CBA.md b/assets/src/CBA/CBA.md index 62bd55b9..fe2e9945 100644 --- a/assets/src/CBA/CBA.md +++ b/assets/src/CBA/CBA.md @@ -15,3 +15,4 @@ CBA(Classification base of Association)算法是一个基于关联规则进 CBA算法作为分类算法,它的判断依据是Apriori算法挖掘出的频繁项。如果一个项集中包含预先知道的属性,同时也包含分类属性值,然后计算该频繁项,能否计算出由已知属性推出决策属性的关联规则,如果满足规则的最小置信度的要求,那么可以把频繁项集中的决策属性值作为最后的分类结果。具体的算法细节如下: 1.输入数据记录; +2.对属性值作数字替换,形成类似关联规则算法的事务记录; From 613490def4d32a14694ee05cfbc6dfad7076014c Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Thu, 4 Jan 2018 09:19:34 +0800 Subject: [PATCH 0040/1480] Update CBA.md --- assets/src/CBA/CBA.md | 1 + 1 file changed, 1 insertion(+) diff --git a/assets/src/CBA/CBA.md b/assets/src/CBA/CBA.md index fe2e9945..7bcac7ad 100644 --- a/assets/src/CBA/CBA.md +++ b/assets/src/CBA/CBA.md @@ -16,3 +16,4 @@ CBA算法作为分类算法,它的判断依据是Apriori算法挖掘出的频 1.输入数据记录; 2.对属性值作数字替换,形成类似关联规则算法的事务记录; +3.根据转化的事务记录,进行Apriori算法计算,挖掘频繁项集; From 2e02021e44e9e64f05819176b4cbb43e9dd9ec99 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Thu, 4 Jan 2018 09:19:57 +0800 Subject: [PATCH 0041/1480] Update CBA.md --- assets/src/CBA/CBA.md | 1 + 1 file changed, 1 insertion(+) diff --git a/assets/src/CBA/CBA.md b/assets/src/CBA/CBA.md index 7bcac7ad..bd23bc56 100644 --- a/assets/src/CBA/CBA.md +++ b/assets/src/CBA/CBA.md @@ -17,3 +17,4 @@ CBA算法作为分类算法,它的判断依据是Apriori算法挖掘出的频 1.输入数据记录; 2.对属性值作数字替换,形成类似关联规则算法的事务记录; 3.根据转化的事务记录,进行Apriori算法计算,挖掘频繁项集; +4.输入查询的属性值,找出符合条件的频繁项集(包含查询属性和分类决策属性);如果找到这样的规则,则算分类成功,输出结果。 From cdc1fa4a3fbb84e05b1c5c899239ea9f8d3e859d Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Thu, 4 Jan 2018 09:20:09 +0800 Subject: [PATCH 0042/1480] Update CBA.md --- assets/src/CBA/CBA.md | 1 + 1 file changed, 1 insertion(+) diff --git a/assets/src/CBA/CBA.md b/assets/src/CBA/CBA.md index bd23bc56..b285a6d2 100644 --- a/assets/src/CBA/CBA.md +++ b/assets/src/CBA/CBA.md @@ -18,3 +18,4 @@ CBA算法作为分类算法,它的判断依据是Apriori算法挖掘出的频 2.对属性值作数字替换,形成类似关联规则算法的事务记录; 3.根据转化的事务记录,进行Apriori算法计算,挖掘频繁项集; 4.输入查询的属性值,找出符合条件的频繁项集(包含查询属性和分类决策属性);如果找到这样的规则,则算分类成功,输出结果。 +Apriori使用逐层搜索的迭代方法。首先,通过扫描事务集,累计每个项的计数,并收集满足最小支持度的项,找出频繁1项集的集合,记为L1。然后用L1寻找频繁2项集的集合L2,依次类推,直到不能再找到频繁k项集。 From 9e766c6bf1845de190b691094fcf25dddcb627ae Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Thu, 4 Jan 2018 09:20:32 +0800 Subject: [PATCH 0043/1480] Update CBA.md --- assets/src/CBA/CBA.md | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/assets/src/CBA/CBA.md b/assets/src/CBA/CBA.md index b285a6d2..0b347455 100644 --- a/assets/src/CBA/CBA.md +++ b/assets/src/CBA/CBA.md @@ -19,3 +19,7 @@ CBA算法作为分类算法,它的判断依据是Apriori算法挖掘出的频 3.根据转化的事务记录,进行Apriori算法计算,挖掘频繁项集; 4.输入查询的属性值,找出符合条件的频繁项集(包含查询属性和分类决策属性);如果找到这样的规则,则算分类成功,输出结果。 Apriori使用逐层搜索的迭代方法。首先,通过扫描事务集,累计每个项的计数,并收集满足最小支持度的项,找出频繁1项集的集合,记为L1。然后用L1寻找频繁2项集的集合L2,依次类推,直到不能再找到频繁k项集。 + + +#### 算法背景 +CBA算法是基于Apriori算法基础上,由Liu,Hsu和MA提出来的。主要是对已经挖掘出的关联规则,做分类判断,所以在某种程度上说CBA算法也是一种集成的挖掘算法。 From b41486e242ffcf92c05a41d0ff561af598407030 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Thu, 4 Jan 2018 09:20:43 +0800 Subject: [PATCH 0044/1480] Update CBA.md --- assets/src/CBA/CBA.md | 2 ++ 1 file changed, 2 insertions(+) diff --git a/assets/src/CBA/CBA.md b/assets/src/CBA/CBA.md index 0b347455..bff38536 100644 --- a/assets/src/CBA/CBA.md +++ b/assets/src/CBA/CBA.md @@ -22,4 +22,6 @@ Apriori使用逐层搜索的迭代方法。首先,通过扫描事务集,累 #### 算法背景 + CBA算法是基于Apriori算法基础上,由Liu,Hsu和MA提出来的。主要是对已经挖掘出的关联规则,做分类判断,所以在某种程度上说CBA算法也是一种集成的挖掘算法。 + From 672ce15ff73a8221463fc240dbe192a20e83f613 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Thu, 4 Jan 2018 09:21:09 +0800 Subject: [PATCH 0045/1480] Update CBA.md --- assets/src/CBA/CBA.md | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/assets/src/CBA/CBA.md b/assets/src/CBA/CBA.md index bff38536..80078b42 100644 --- a/assets/src/CBA/CBA.md +++ b/assets/src/CBA/CBA.md @@ -25,3 +25,8 @@ Apriori使用逐层搜索的迭代方法。首先,通过扫描事务集,累 CBA算法是基于Apriori算法基础上,由Liu,Hsu和MA提出来的。主要是对已经挖掘出的关联规则,做分类判断,所以在某种程度上说CBA算法也是一种集成的挖掘算法。 +#### 相关应用 + +CBA算法应用广泛,如可用在保险领域、生物学领域、地震研究等领域中。 + +如可用于消费市场价格分析,猜测顾客的消费习惯;网络安全领域中的入侵检测技术;可用在用于高校管理中,根据挖掘规则可以有效地辅助学校管理部门有针对性的开展贫困助学工作;也可用在移动通信领域中,指导运营商的业务运营和辅助业务提供商的决策制定等。 From cde65b7ce75b1b22099bf1e3c4beffc78567d3f4 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Thu, 4 Jan 2018 09:21:52 +0800 Subject: [PATCH 0046/1480] Update CBA.md --- assets/src/CBA/CBA.md | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/assets/src/CBA/CBA.md b/assets/src/CBA/CBA.md index 80078b42..6d788700 100644 --- a/assets/src/CBA/CBA.md +++ b/assets/src/CBA/CBA.md @@ -30,3 +30,7 @@ CBA算法是基于Apriori算法基础上,由Liu,Hsu和MA提出来的。主要 CBA算法应用广泛,如可用在保险领域、生物学领域、地震研究等领域中。 如可用于消费市场价格分析,猜测顾客的消费习惯;网络安全领域中的入侵检测技术;可用在用于高校管理中,根据挖掘规则可以有效地辅助学校管理部门有针对性的开展贫困助学工作;也可用在移动通信领域中,指导运营商的业务运营和辅助业务提供商的决策制定等。 + +#### 优点 + +优点:分类的准确度较高,在大量数据集上比C4.5更精确。 From cb6717a397f3936b299645ae955131524e664141 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Thu, 4 Jan 2018 09:24:29 +0800 Subject: [PATCH 0047/1480] change page of RAM --- assets/src/RAM/RAM.0.1.md | 5 +++++ 1 file changed, 5 insertions(+) create mode 100644 assets/src/RAM/RAM.0.1.md diff --git a/assets/src/RAM/RAM.0.1.md b/assets/src/RAM/RAM.0.1.md new file mode 100644 index 00000000..c0b01b83 --- /dev/null +++ b/assets/src/RAM/RAM.0.1.md @@ -0,0 +1,5 @@ +### Deeplearning Algorithms tutorial +谷歌的人工智能位于全球前列,在图像识别、语音识别、无人驾驶等技术上都已经落地。而百度实质意义上扛起了国内的人工智能的大旗,覆盖无人驾驶、智能助手、图像识别等许多层面。苹果业已开始全面拥抱机器学习,新产品进军家庭智能音箱并打造工作站级别Mac。另外,腾讯的深度学习平台Mariana已支持了微信语音识别的语音输入法、语音开放平台、长按语音消息转文本等产品,在微信图像识别中开始应用。全球前十大科技公司全部发力人工智能理论研究和应用的实现,虽然入门艰难,但是一旦入门,高手也就在你的不远处! +AI的开发离不开算法那我们就接下来开始学习算法吧! + +#### 回归算法 From 798b34a64afbde2e437bc62701cc09babdfd6bd9 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Thu, 4 Jan 2018 09:25:35 +0800 Subject: [PATCH 0048/1480] change page of RAM --- assets/src/RAM/RAM.0.1.md | 9 +++++++++ 1 file changed, 9 insertions(+) diff --git a/assets/src/RAM/RAM.0.1.md b/assets/src/RAM/RAM.0.1.md index c0b01b83..e65cd4ed 100644 --- a/assets/src/RAM/RAM.0.1.md +++ b/assets/src/RAM/RAM.0.1.md @@ -3,3 +3,12 @@ AI的开发离不开算法那我们就接下来开始学习算法吧! #### 回归算法 + +回归方法是对数值型连续随机变量进行预测和建模的监督学习算法。其特点是标注的数据集具有数值型的目标变量。 + +常用的回归方法包括: +线性回归:使用超平面拟合数据集 +最近邻算法:通过搜寻最相似的训练样本来预测新样本的值 +决策树和回归树:将数据集分割为不同分支而实现分层学习 +集成方法:组合多个弱学习算法构造一种强学习算法,如随机森林(RF)和梯度提升树(GBM)等 +深度学习:使用多层神经网络学习复杂模型 \ No newline at end of file From 10b6b85890eb797a7e784601298b40eb0e3d8795 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Thu, 4 Jan 2018 09:27:52 +0800 Subject: [PATCH 0049/1480] change ARAM --- assets/src/RAM/RAM.0.1.md | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) diff --git a/assets/src/RAM/RAM.0.1.md b/assets/src/RAM/RAM.0.1.md index e65cd4ed..6012ddb6 100644 --- a/assets/src/RAM/RAM.0.1.md +++ b/assets/src/RAM/RAM.0.1.md @@ -7,8 +7,8 @@ AI的开发离不开算法那我们就接下来开始学习算法吧! 回归方法是对数值型连续随机变量进行预测和建模的监督学习算法。其特点是标注的数据集具有数值型的目标变量。 常用的回归方法包括: -线性回归:使用超平面拟合数据集 -最近邻算法:通过搜寻最相似的训练样本来预测新样本的值 -决策树和回归树:将数据集分割为不同分支而实现分层学习 -集成方法:组合多个弱学习算法构造一种强学习算法,如随机森林(RF)和梯度提升树(GBM)等 -深度学习:使用多层神经网络学习复杂模型 \ No newline at end of file +* 线性回归:使用超平面拟合数据集 +* 最近邻算法:通过搜寻最相似的训练样本来预测新样本的值 +* 决策树和回归树:将数据集分割为不同分支而实现分层学习 +* 集成方法:组合多个弱学习算法构造一种强学习算法,如随机森林(RF)和梯度提升树(GBM)等 +* 深度学习:使用多层神经网络学习复杂模型 \ No newline at end of file From 201893c2c3bd9754e60f95967a73ee35e6acce6f Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Thu, 4 Jan 2018 09:33:31 +0800 Subject: [PATCH 0050/1480] change page --- README.md | 2 +- assets/src/RAM/RAM.0.1.md | 22 ++++++++++++++++++++-- 2 files changed, 21 insertions(+), 3 deletions(-) diff --git a/README.md b/README.md index 8a458a5d..7a8c3b1c 100644 --- a/README.md +++ b/README.md @@ -66,7 +66,7 @@ 后面的算法和我们的算法模型,我会持续更新整理,后续的算法章节会不断的补上,希望可以对新入门学习AI开发和算法的开发者有所帮助! #### 算法模型 - * [回归算法]() + * [回归算法](https://github.com/KeKe-Li/tutorial/blob/master/assets/src/RAM/RAM.0.1.md) * [线性回归]() * [逻辑回归]() * [多元自适应回归(MARS)]() diff --git a/assets/src/RAM/RAM.0.1.md b/assets/src/RAM/RAM.0.1.md index 6012ddb6..1d1d0311 100644 --- a/assets/src/RAM/RAM.0.1.md +++ b/assets/src/RAM/RAM.0.1.md @@ -4,11 +4,29 @@ AI的开发离不开算法那我们就接下来开始学习算法吧! #### 回归算法 -回归方法是对数值型连续随机变量进行预测和建模的监督学习算法。其特点是标注的数据集具有数值型的目标变量。 +回归方法是对数值型连续随机变量进行预测和建模的监督学习算法。其特点是标注的数据集具有数值型的目标变量。回归的目的是预测数值型的目标值。 +最直接的办法是依据输入写出一个目标值的计算公式,该公式就是所谓的回归方程(regression equation)。求回归方程中的回归系数的过程就是回归。 常用的回归方法包括: * 线性回归:使用超平面拟合数据集 * 最近邻算法:通过搜寻最相似的训练样本来预测新样本的值 * 决策树和回归树:将数据集分割为不同分支而实现分层学习 * 集成方法:组合多个弱学习算法构造一种强学习算法,如随机森林(RF)和梯度提升树(GBM)等 -* 深度学习:使用多层神经网络学习复杂模型 \ No newline at end of file +* 深度学习:使用多层神经网络学习复杂模型 + + +#### 如何应用 + +* 收集数据:可以使用任何方法。 +* 准备数据:回归需要数值型数据,标称型数据将被转换成二值型数据。 +* 分析数据:绘出数据的可视化二维图将有助于对数据做出理解和分析,在采用缩减法求得新回归系数之后,可以将新拟合线绘在图上作为对比。 +* 训练算法:找到回归系数。 +* 测试算法:使用 R2 或者预测值和数据的拟合度,来分析模型的效果。 +* 使用算法:使用回归,可以在给定输入的时候预测出一个数值,这是对分类方法的提升,因为这样可以预测连续型数据而不仅仅是离散的类别标签。 + + +#### 优缺点 + +* 优点:结果容易理解,计算上不复杂。 +* 缺点:对非线性的数据拟合不好。 +* 适用数据范围:数值型和标称型。 \ No newline at end of file From c6d2edf590c7b0880288a55ad6d64b62a08c56af Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Thu, 4 Jan 2018 09:37:46 +0800 Subject: [PATCH 0051/1480] change page RAM --- assets/src/RAM/RAM.0.2.md | 19 +++++++++++++++++++ 1 file changed, 19 insertions(+) create mode 100644 assets/src/RAM/RAM.0.2.md diff --git a/assets/src/RAM/RAM.0.2.md b/assets/src/RAM/RAM.0.2.md new file mode 100644 index 00000000..a9e87a2e --- /dev/null +++ b/assets/src/RAM/RAM.0.2.md @@ -0,0 +1,19 @@ +### Deeplearning Algorithms tutorial +谷歌的人工智能位于全球前列,在图像识别、语音识别、无人驾驶等技术上都已经落地。而百度实质意义上扛起了国内的人工智能的大旗,覆盖无人驾驶、智能助手、图像识别等许多层面。苹果业已开始全面拥抱机器学习,新产品进军家庭智能音箱并打造工作站级别Mac。另外,腾讯的深度学习平台Mariana已支持了微信语音识别的语音输入法、语音开放平台、长按语音消息转文本等产品,在微信图像识别中开始应用。全球前十大科技公司全部发力人工智能理论研究和应用的实现,虽然入门艰难,但是一旦入门,高手也就在你的不远处! +AI的开发离不开算法那我们就接下来开始学习算法吧! +回归方法是对数值型连续随机变量进行预测和建模的监督学习算法。其特点是标注的数据集具有数值型的目标变量。回归的目的是预测数值型的目标值。 + + +常用的回归方法包括: +* 线性回归:使用超平面拟合数据集 +* 最近邻算法:通过搜寻最相似的训练样本来预测新样本的值 +* 决策树和回归树:将数据集分割为不同分支而实现分层学习 +* 集成方法:组合多个弱学习算法构造一种强学习算法,如随机森林(RF)和梯度提升树(GBM)等 +* 深度学习:使用多层神经网络学习复杂模型 + +#### 线性回归 + +线性回归是最简单的回归方法,它的目标是使用超平面拟合数据集,即学习一个线性模型以尽可能准确的预测实值输出标记。 +线性回归是利用数理统计中回归分析,来确定两种或两种以上变量间相互依赖的定量关系的一种统计分析方法,运用十分广泛。其表达形式为y = w'x+e,e为误差服从均值为0的正态分布。 +回归分析中,只包括一个自变量和一个因变量,且二者的关系可用一条直线近似表示,这种回归分析称为一元线性回归分析。如果回归分析中包括两个或两个以上的自变量,且因变量和自变量之间是线性关系,则称为多元线性回归分析。 + From fe280955c97f51dfc98c3cd0ca4b14f68c554405 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Thu, 4 Jan 2018 09:38:34 +0800 Subject: [PATCH 0052/1480] Update README.md --- README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/README.md b/README.md index 7a8c3b1c..736ba1b4 100644 --- a/README.md +++ b/README.md @@ -67,7 +67,7 @@ #### 算法模型 * [回归算法](https://github.com/KeKe-Li/tutorial/blob/master/assets/src/RAM/RAM.0.1.md) - * [线性回归]() + * [线性回归](https://github.com/KeKe-Li/tutorial/blob/master/assets/src/RAM/RAM.0.2.md) * [逻辑回归]() * [多元自适应回归(MARS)]() * [本地散点平滑估计(LOESS)]() From 13f3fe4500a311bf60e0e0e522d81b2e149bf134 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Thu, 4 Jan 2018 15:57:38 +0800 Subject: [PATCH 0053/1480] Update README.md --- README.md | 1 + 1 file changed, 1 insertion(+) diff --git a/README.md b/README.md index 736ba1b4..c2d39ef9 100644 --- a/README.md +++ b/README.md @@ -163,6 +163,7 @@ * [梯度下降](https://github.com/KeKe-Li/book/blob/master/AI/%E6%A2%AF%E5%BA%A6%E4%B8%8B%E9%99%8D.pdf) * [无监督神经元](https://blog.openai.com/unsupervised-sentiment-neuron/) * [Tensorflow实践](https://tensorflow.feisky.xyz/install/cpu.html) +* [Artificial Intelligence]() ### 机器学习 觉得此文章不错可以给我star!如果有问题可以加我的微信Sen0676,可以一起交流算法和机器学习! From 95f80f02ba4d5dbc9899913b4e88b9bb4807700b Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Thu, 4 Jan 2018 15:57:54 +0800 Subject: [PATCH 0054/1480] Update README.md --- README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/README.md b/README.md index c2d39ef9..05b75ad3 100644 --- a/README.md +++ b/README.md @@ -163,7 +163,7 @@ * [梯度下降](https://github.com/KeKe-Li/book/blob/master/AI/%E6%A2%AF%E5%BA%A6%E4%B8%8B%E9%99%8D.pdf) * [无监督神经元](https://blog.openai.com/unsupervised-sentiment-neuron/) * [Tensorflow实践](https://tensorflow.feisky.xyz/install/cpu.html) -* [Artificial Intelligence]() +* [Artificial Intelligence](https://github.com/owainlewis/awesome-artificial-intelligence) ### 机器学习 觉得此文章不错可以给我star!如果有问题可以加我的微信Sen0676,可以一起交流算法和机器学习! From b23423ca8534a8f2a193abfa2038a1c4edee8801 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Fri, 5 Jan 2018 09:10:08 +0800 Subject: [PATCH 0055/1480] change page of KNN --- assets/src/KNN/KNN.md | 11 +++++++++++ 1 file changed, 11 insertions(+) create mode 100644 assets/src/KNN/KNN.md diff --git a/assets/src/KNN/KNN.md b/assets/src/KNN/KNN.md new file mode 100644 index 00000000..8d93978d --- /dev/null +++ b/assets/src/KNN/KNN.md @@ -0,0 +1,11 @@ +### Deeplearning Algorithms tutorial +谷歌的人工智能位于全球前列,在图像识别、语音识别、无人驾驶等技术上都已经落地。而百度实质意义上扛起了国内的人工智能的大旗,覆盖无人驾驶、智能助手、图像识别等许多层面。苹果业已开始全面拥抱机器学习,新产品进军家庭智能音箱并打造工作站级别Mac。另外,腾讯的深度学习平台Mariana已支持了微信语音识别的语音输入法、语音开放平台、长按语音消息转文本等产品,在微信图像识别中开始应用。全球前十大科技公司全部发力人工智能理论研究和应用的实现,虽然入门艰难,但是一旦入门,高手也就在你的不远处! +AI的开发离不开算法那我们就接下来开始学习算法吧! + +机器学习是一门多领域交叉学科,涉及概率论、统计学、逼近论、凸分析、算法复杂度理论等多门学科。主要研究计算机怎样模拟或实现人类的学习行为,以获取新的知识和技能,重新组织已有的知识结构,不断的改善自身的性能。 + +机器学习理论主要是设计和分析一些让计算机可以自动“学习”的算法。这些算法是一类能从数据中自动分析获得规律,并利用规律对未知数据进行预测的算法。简而言之,机器学习主要以数据为基础,通过大数据本身,运用计算机自我学习来寻找数据本身的规律,而这是机器学习与统计分析的基本区别。 + +机器学习主要有三种方式:监督学习,无监督学习与半监督学习。 + +#### KNN From 4cdc47ca66970b3e53e43f9232f1259301bcc443 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Fri, 5 Jan 2018 09:11:01 +0800 Subject: [PATCH 0056/1480] Update README.md --- README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/README.md b/README.md index 05b75ad3..9a74564c 100644 --- a/README.md +++ b/README.md @@ -56,7 +56,7 @@ * [PageRank](https://github.com/KeKe-Li/tutorial/blob/master/assets/src/PRK/PRK.md) * [AdaBoost](https://github.com/KeKe-Li/tutorial/blob/master/assets/src/ABT/ABT.md) * [CBA](https://github.com/KeKe-Li/tutorial/blob/master/assets/src/CBA/CBA.md) - * [KNN]() + * [KNN](https://github.com/KeKe-Li/tutorial/blob/master/assets/src/KNN/KNN.md) * [Hopfield神经网络]() * [决策树]() * [聚类分析]() From cd22708b098a2f842b32fa5b2d6c84616b1b175a Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Fri, 5 Jan 2018 09:12:16 +0800 Subject: [PATCH 0057/1480] Update KNN.md --- assets/src/KNN/KNN.md | 1 + 1 file changed, 1 insertion(+) diff --git a/assets/src/KNN/KNN.md b/assets/src/KNN/KNN.md index 8d93978d..15a7f66c 100644 --- a/assets/src/KNN/KNN.md +++ b/assets/src/KNN/KNN.md @@ -9,3 +9,4 @@ AI的开发离不开算法那我们就接下来开始学习算法吧! 机器学习主要有三种方式:监督学习,无监督学习与半监督学习。 #### KNN +K最近邻算法是一种基于类比的分类方法,主要通过给定的检验组与和它相似的训练组进行比较来学习。训练组用n个属性来描述,每个元组代表n维空间上的点。当给定一个未知元组时,K最近邻分类法搜索该模式空间,找出最接近未知元组的k个训练组,并将未知元组指派到模式空间中它的k个最近邻中的多数类中。 From 2aebcb3f4612f0c7c9f7be82765960a78698be28 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Fri, 5 Jan 2018 09:13:45 +0800 Subject: [PATCH 0058/1480] change page --- assets/images/87.jpg | Bin 0 -> 1365 bytes assets/images/88.jpg | Bin 0 -> 767 bytes assets/images/89.jpg | Bin 0 -> 570 bytes assets/images/90.jpg | Bin 0 -> 479 bytes 4 files changed, 0 insertions(+), 0 deletions(-) create mode 100644 assets/images/87.jpg create mode 100644 assets/images/88.jpg create mode 100644 assets/images/89.jpg create mode 100644 assets/images/90.jpg diff --git a/assets/images/87.jpg b/assets/images/87.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f0524a1650f424046900a1bc302d9b84d48ceb5d GIT binary patch literal 1365 zcmV-b1*-aqP)
![]()
00001b5ch_0Itp) z=>Px#fly3TMF0Q*0000006G8wW&k=m06Jy>I* YMklVJ9+uM-a|F&iTwq`oEkdU^v z|F-|Ow*T9<|Nq;NI@`8p+uM-a|NsBCX8+re|NplC|JxJ=cHRI000DGTPE!Ct=GbNc z000SaNLh0L01m_e01m_fl`9S#0000PbVXQnQ*UN;cVTj608MFQa&L2QW^^D=W@c$) zWq<88aR2}WJ4r-ARA@u(T8VPgFc7slniNP{j<}^Y -PfrWvEf%!2|<|B 3- kMB(>cmQ|~4PBvWuq-LFaUfQf-vfRPHm)RIDz=5;NbJtC~Vwbgd z&smA*I+MX|S((3b+ZF;ptD)Mmb~~ronm8fdeugeOtHT5^rY!>(otUvwG#oTaInUd+ ztn`!1AO*(?)wyvK@R&55O5VIt-`q- `UPPcclOA(++t0R}aE_x4IyaMAr;nqs-#g1?l3&qNT@Qbat!tnrLW zKto0e+r)FlsK_T{&nI1!g|kv5^4l`vb X1zZW8J4R!WF;&nA!Qwp)Hyf-FtP08Z129B$ZQWFp4I2sXSX&U z;6W8a&N4W~I8J*?Nx8)PIF%i<4)rL8Af?n<%+8ubEN#fc8E3ID%qbqaf}N9A?|pM& zq^K%TVy+_*uO*rWEQI+4*Mm)fGbynWvkDdg?e+~OR1RZi)+S;p-nwok*|=r-;i=d( zpOx-Df9_Z+^Qr>>jL}iQKvxf+H@`XbqWOZO5+M3fX8}rSIPJ}Gw zW*H-GMTHyc?t9EiQ5EfnnSH^I{}ve{;l!@kA8e44*}67|L0%XkZKT@|WyuwAGb|<& zZ=Me=5p~5b!mCPI#kMOziIfi$yb(_M1uAWZ#U$b#ngTaUmSUWU<$lGMnNmto8}L~s zyS@MMeZLvdW>`!lzIP?8lcJ=j6R~86RDjSBQ;J^7WtnuR6S;RT7+^?DCSKg}aRR0- ztA<$ZaXO~(lx|n#lrn|g9=`|Gu7tTk%DTrZ9O+4gGfXKZW@_&X8VAJ5k)Z5V%WYX5 zT+zTgBz}tG8C+TjmCqEB`Fo c9+cTNw)#rh|^{QgQO7i`TCB zhlTZvwp(-kc5&wkDW00z-mBoC%=zyNu+bQzLa+b!!!L7^Ro7@+KTI(u6S)RIY|SZ3 zL-x3iFELiHp047{eeL$`$BWL2L%l@0t{uA0)a3>@ol^%Q?eAK?#3=Ew{C#0yI0OFx XGtpz&8l?##00000NkvXXu0mjf4C;PS literal 0 HcmV?d00001 diff --git a/assets/images/88.jpg b/assets/images/88.jpg new file mode 100644 index 0000000000000000000000000000000000000000..870e42478f9568fd834593f0b4880a40f7499421 GIT binary patch literal 767 zcmeAS@N?(olHy`uVBq!ia0vp^bwI4e!3-pw9?3@oDaPU;cPEB*=VV?2In@C^A+A6g z46GO!(ip6OXaYmpHXz4}!ODsu&B|(mRa%-AP-Macs|mN$po)M(R%uof(tsk{wxw 4nJ@ErkR#;MwT(m+AU64!{5;QX|b^2DN42H(u&qWoeY zPr*MeEi*Z_{;g>t(3IDnE{-7)?r*2pM;}rUIKJ?cd+O$>+*v-Vtl$5|C*Dsse%@p$ z7;9Vk-$$}w`+M2%H3@IB!dVY#Z;(`fXr8-&v0T@;$^TFI9IbwB^VBRhGlIXOB|ZPF zk({B_&7P0SQ%>%CvsO)M^N#a|e?- T^G)*fa_?LG^5Ibr z75pEX?H2TTxPRaAIY}(qO_6MxQ)B(Kw%HuDF7i;^kfic}bHU>!@^7-F&dz6j*(adk zB4ubKqG%eS`9g8(hOIyKX3y+bYj-xBzm(HND %;dM6PN@XBVKM%xD$Ig-L~U_B#V9Ql8L $VAd1a4(dJiA-msakha i<#&+S+U|!Po$*VjT7TBK?sN{6ay?!BT-G@yGywpUY*b(X literal 0 HcmV?d00001 diff --git a/assets/images/89.jpg b/assets/images/89.jpg new file mode 100644 index 0000000000000000000000000000000000000000..6f79590006d647b1402f2806f999eb1cede1790a GIT binary patch literal 570 zcmeAS@N?(olHy`uVBq!ia0vp^9zblz!3-q#d#qgvq!^2X+?^QKos)S9 ?z>u~L$g*OvvSLWHvYKF(mSzPMnQ%J|ES<&x (vv)Z-|DE$BS1gqN<(r(`d3g5o{|36R;C IH^e@BDuOFYQyy^&Bvu#zvq|(Gx2Q}lzuh(7SnGoF| zz;Pgj;p)6w$B+Gc?C8B!`=GD)#f$~h8N}uXNHEErY^aHAnB7`_<8RX9yiZ1E)44fP zcJeMuJL=%UGf9MNzV;o1n*C?WrtxYmHce%f%+PxI_rs}<-wRYEPsCpf_I$Xieb=r1 z5w`<^u3z+ESTiY!Eju(wL}2|AjTuvH7?`ad_2ot0dmgFZzV6nexv#_T$1mO5Xlk=q zxOVH(S+^FLK4v$Hd38doNWnV(gx3t6PoYN6`*#}jNpjX}L|XScEjpI6-{|dzlpnP+ qOj};4W}G{=*v ?X%lY$- 9#z- zxjnwxd;T}CYuI$8Leu-t<1!&b! Id3ucebwE@{H@%xp(i1R;nnb-i$YRb;L%yR?p_vbJYy~ znbPm)UHlYU5fWHG^UzM0ZK00?qjz5XxZq}Zh{NF*0nx^5`!28_H9ntpNB>61Cf}c| zEq|TX+%t%0?g|Nc5_V<1U#jZPgslReD)u`bpQ~Zc`N*9fzEEoyFaQ`lUHx3vIVCg! E06mqDate: Fri, 5 Jan 2018 09:15:27 +0800 Subject: [PATCH 0059/1480] Update KNN.md --- assets/src/KNN/KNN.md | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/assets/src/KNN/KNN.md b/assets/src/KNN/KNN.md index 15a7f66c..b5c7f1ee 100644 --- a/assets/src/KNN/KNN.md +++ b/assets/src/KNN/KNN.md @@ -10,3 +10,9 @@ AI的开发离不开算法那我们就接下来开始学习算法吧! #### KNN K最近邻算法是一种基于类比的分类方法,主要通过给定的检验组与和它相似的训练组进行比较来学习。训练组用n个属性来描述,每个元组代表n维空间上的点。当给定一个未知元组时,K最近邻分类法搜索该模式空间,找出最接近未知元组的k个训练组,并将未知元组指派到模式空间中它的k个最近邻中的多数类中。 + +其中“最近邻”主要是以距离来度量的,一般使用欧几里得距离度量两个点或元组的距离,也可以使用曼哈顿距离或其他距离;欧几里得距离的主要计算公式如下: + +
+ From 3da024a33c8f46261b62c74fac22dae8d426889f Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Fri, 5 Jan 2018 09:16:28 +0800 Subject: [PATCH 0060/1480] Update KNN.md --- assets/src/KNN/KNN.md | 3 +++ 1 file changed, 3 insertions(+) diff --git a/assets/src/KNN/KNN.md b/assets/src/KNN/KNN.md index b5c7f1ee..69491284 100644 --- a/assets/src/KNN/KNN.md +++ b/assets/src/KNN/KNN.md @@ -16,3 +16,6 @@ K最近邻算法是一种基于类比的分类方法,主要通过给定的检+
+说明:为了防止具有较大初始值域的属性比较小初始值域的属性的权重过大,在计算距离之前,需对每个属性值进行规范化。一般的规划方法有最小-最大规范化,零均值规范化,小数定标规范化等。 + +最小-最大规范化:将原始数据值映射到[0,1]空间中,假定minA和maxA分别是属性A的最小值和最大值,则规范化的公式为: From 89715c34d10107611b389f3d26d214d2af45cf8c Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Fri, 5 Jan 2018 09:16:48 +0800 Subject: [PATCH 0061/1480] Update KNN.md --- assets/src/KNN/KNN.md | 3 +++ 1 file changed, 3 insertions(+) diff --git a/assets/src/KNN/KNN.md b/assets/src/KNN/KNN.md index 69491284..d8e5f005 100644 --- a/assets/src/KNN/KNN.md +++ b/assets/src/KNN/KNN.md @@ -19,3 +19,6 @@ K最近邻算法是一种基于类比的分类方法,主要通过给定的检 说明:为了防止具有较大初始值域的属性比较小初始值域的属性的权重过大,在计算距离之前,需对每个属性值进行规范化。一般的规划方法有最小-最大规范化,零均值规范化,小数定标规范化等。 最小-最大规范化:将原始数据值映射到[0,1]空间中,假定minA和maxA分别是属性A的最小值和最大值,则规范化的公式为: +
+
From 2d66c1c0f30da90d0904aea855102e2534de5df5 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Fri, 5 Jan 2018 09:21:15 +0800 Subject: [PATCH 0062/1480] Update KNN.md --- assets/src/KNN/KNN.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/assets/src/KNN/KNN.md b/assets/src/KNN/KNN.md index d8e5f005..5015ceec 100644 --- a/assets/src/KNN/KNN.md +++ b/assets/src/KNN/KNN.md @@ -20,5 +20,5 @@ K最近邻算法是一种基于类比的分类方法,主要通过给定的检 最小-最大规范化:将原始数据值映射到[0,1]空间中,假定minA和maxA分别是属性A的最小值和最大值,则规范化的公式为:+
-
From 8d0afbdbe9a2bbd1189fb8e8f564f09f24bb94c0 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Fri, 5 Jan 2018 09:21:50 +0800 Subject: [PATCH 0063/1480] Update KNN.md --- assets/src/KNN/KNN.md | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) diff --git a/assets/src/KNN/KNN.md b/assets/src/KNN/KNN.md index 5015ceec..fc99ed5d 100644 --- a/assets/src/KNN/KNN.md +++ b/assets/src/KNN/KNN.md @@ -20,5 +20,9 @@ K最近邻算法是一种基于类比的分类方法,主要通过给定的检 最小-最大规范化:将原始数据值映射到[0,1]空间中,假定minA和maxA分别是属性A的最小值和最大值,则规范化的公式为:+
![]()
-
+零均值规范化:基于属性A的均值和标准差上的规范化方法,具体计算如下: ++
+
+
From ec26706ae52a3bd6317beb75aee9a8ea87dbe1fa Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Fri, 5 Jan 2018 09:22:01 +0800 Subject: [PATCH 0064/1480] Update KNN.md --- assets/src/KNN/KNN.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/assets/src/KNN/KNN.md b/assets/src/KNN/KNN.md index fc99ed5d..0ba8d67c 100644 --- a/assets/src/KNN/KNN.md +++ b/assets/src/KNN/KNN.md @@ -24,5 +24,5 @@ K最近邻算法是一种基于类比的分类方法,主要通过给定的检 零均值规范化:基于属性A的均值和标准差上的规范化方法,具体计算如下:![]()
-
From 74217c5c514a2ad7fa26e7f30810fe93e9c5a699 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Fri, 5 Jan 2018 09:22:24 +0800 Subject: [PATCH 0065/1480] Update KNN.md --- assets/src/KNN/KNN.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/assets/src/KNN/KNN.md b/assets/src/KNN/KNN.md index 0ba8d67c..81d8e16c 100644 --- a/assets/src/KNN/KNN.md +++ b/assets/src/KNN/KNN.md @@ -16,7 +16,7 @@ K最近邻算法是一种基于类比的分类方法,主要通过给定的检+
![]()
-说明:为了防止具有较大初始值域的属性比较小初始值域的属性的权重过大,在计算距离之前,需对每个属性值进行规范化。一般的规划方法有最小-最大规范化,零均值规范化,小数定标规范化等。 +注意:为了防止具有较大初始值域的属性比较小初始值域的属性的权重过大,在计算距离之前,需对每个属性值进行规范化。一般的规划方法有最小-最大规范化,零均值规范化,小数定标规范化等。 最小-最大规范化:将原始数据值映射到[0,1]空间中,假定minA和maxA分别是属性A的最小值和最大值,则规范化的公式为:
From 6b8d15477e1bf66d0ed4948f1c424d11019365ce Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Fri, 5 Jan 2018 09:22:40 +0800 Subject: [PATCH 0066/1480] Update KNN.md --- assets/src/KNN/KNN.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/assets/src/KNN/KNN.md b/assets/src/KNN/KNN.md index 81d8e16c..2300eff0 100644 --- a/assets/src/KNN/KNN.md +++ b/assets/src/KNN/KNN.md @@ -24,5 +24,5 @@ K最近邻算法是一种基于类比的分类方法,主要通过给定的检
零均值规范化:基于属性A的均值和标准差上的规范化方法,具体计算如下:-
From fa0d04713f97c6f8c41d85c1b450615177872c7f Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Fri, 5 Jan 2018 09:23:20 +0800 Subject: [PATCH 0067/1480] Update KNN.md --- assets/src/KNN/KNN.md | 2 ++ 1 file changed, 2 insertions(+) diff --git a/assets/src/KNN/KNN.md b/assets/src/KNN/KNN.md index 2300eff0..8dffdc4b 100644 --- a/assets/src/KNN/KNN.md +++ b/assets/src/KNN/KNN.md @@ -26,3 +26,5 @@ K最近邻算法是一种基于类比的分类方法,主要通过给定的检+
![]()
+ +小数定标规范化:通过移动属性A的小数点位置进行规范化,小数点的移动位数依赖于A的最大绝对值。具体计算如下:(其中j是使得MAX( v’ )<1的最小整数) From 92d9d7071dae0919bf4d38043506f5fb916837a6 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Fri, 5 Jan 2018 09:24:35 +0800 Subject: [PATCH 0068/1480] Update KNN.md --- assets/src/KNN/KNN.md | 3 +++ 1 file changed, 3 insertions(+) diff --git a/assets/src/KNN/KNN.md b/assets/src/KNN/KNN.md index 8dffdc4b..a90221e4 100644 --- a/assets/src/KNN/KNN.md +++ b/assets/src/KNN/KNN.md @@ -28,3 +28,6 @@ K最近邻算法是一种基于类比的分类方法,主要通过给定的检 小数定标规范化:通过移动属性A的小数点位置进行规范化,小数点的移动位数依赖于A的最大绝对值。具体计算如下:(其中j是使得MAX( v’ )<1的最小整数) +
![]()
+
From 3c72227cf2e680798d67fa879f9a1586295c5292 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Fri, 5 Jan 2018 09:24:47 +0800 Subject: [PATCH 0069/1480] Update KNN.md --- assets/src/KNN/KNN.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/assets/src/KNN/KNN.md b/assets/src/KNN/KNN.md index a90221e4..495a11ad 100644 --- a/assets/src/KNN/KNN.md +++ b/assets/src/KNN/KNN.md @@ -29,5 +29,5 @@ K最近邻算法是一种基于类比的分类方法,主要通过给定的检 小数定标规范化:通过移动属性A的小数点位置进行规范化,小数点的移动位数依赖于A的最大绝对值。具体计算如下:(其中j是使得MAX( v’ )<1的最小整数)+
-
From 9ce7b7c5f87321bae53a5d65c98108e4c1655f1a Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Fri, 5 Jan 2018 09:25:20 +0800 Subject: [PATCH 0070/1480] Update KNN.md --- assets/src/KNN/KNN.md | 8 ++++++++ 1 file changed, 8 insertions(+) diff --git a/assets/src/KNN/KNN.md b/assets/src/KNN/KNN.md index 495a11ad..a02b248e 100644 --- a/assets/src/KNN/KNN.md +++ b/assets/src/KNN/KNN.md @@ -31,3 +31,11 @@ K最近邻算法是一种基于类比的分类方法,主要通过给定的检+
![]()
+ +最近邻数K的确定,主要原理是选取产生最小误差率的k值。每次从k=1开始,使用检验集估计分类器的误差率;每次都允许增加一个近邻,重复该过程,选择误差率最小的k值。 + +#### 算法背景 + +KNN算法是由Cover和Hart提出来的,是一种懒惰的、有监督的、基于实例的机器学习方法。同时是向量空间模型下最好的分类算法之一。 + + From 85b173ed0f1904a5fd028bf1d4489a76af51917c Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Fri, 5 Jan 2018 09:25:56 +0800 Subject: [PATCH 0071/1480] Update KNN.md --- assets/src/KNN/KNN.md | 3 +++ 1 file changed, 3 insertions(+) diff --git a/assets/src/KNN/KNN.md b/assets/src/KNN/KNN.md index a02b248e..a6f0da1b 100644 --- a/assets/src/KNN/KNN.md +++ b/assets/src/KNN/KNN.md @@ -38,4 +38,7 @@ K最近邻算法是一种基于类比的分类方法,主要通过给定的检 KNN算法是由Cover和Hart提出来的,是一种懒惰的、有监督的、基于实例的机器学习方法。同时是向量空间模型下最好的分类算法之一。 +#### 算法应用 +K最近邻算法是一种基本的分类方法,主要对数据进行分类处理。分类时,对新的记录,根据其K个最邻近的训练记录,这K个记录的多数属于某个类,就把该新的记录分为这个类。K值一般选取比较小的数值。通常采用交叉验证法来选取最优的K值。 + From bae562b307ed806d7ae605a6fafe0bd1d3b291d6 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Fri, 5 Jan 2018 09:26:26 +0800 Subject: [PATCH 0072/1480] Update KNN.md --- assets/src/KNN/KNN.md | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/assets/src/KNN/KNN.md b/assets/src/KNN/KNN.md index a6f0da1b..3978c689 100644 --- a/assets/src/KNN/KNN.md +++ b/assets/src/KNN/KNN.md @@ -42,3 +42,8 @@ KNN算法是由Cover和Hart提出来的,是一种懒惰的、有监督的、 K最近邻算法是一种基本的分类方法,主要对数据进行分类处理。分类时,对新的记录,根据其K个最邻近的训练记录,这K个记录的多数属于某个类,就把该新的记录分为这个类。K值一般选取比较小的数值。通常采用交叉验证法来选取最优的K值。 +#### 优缺点 +优点:简单,有效;重新训练的代价低;计算时间和空间训练集的规模;对于数据集的交叉或重叠较多的待分样本集来说,KNN算法比其他算法合适;比较适用于样本容量较大的类域的自动分类,而对样本容量小的类域会产生较大的误分。 + +缺点:输出的可解释性不强;计算量较大;对数据样本容量相差较大的,应该先进行规范化处理。 + From e4097eeb27a65b254dc1cc4beba571ee25a45961 Mon Sep 17 00:00:00 2001 From: keke <2536485681li@gmail.com> Date: Fri, 5 Jan 2018 09:30:30 +0800 Subject: [PATCH 0073/1480] add HPD --- assets/images/91.jpg | Bin 0 -> 16435 bytes assets/images/92.jpg | Bin 0 -> 27482 bytes assets/images/93.jpg | Bin 0 -> 4994 bytes assets/images/94.jpg | Bin 0 -> 2809 bytes assets/images/95.jpg | Bin 0 -> 7754 bytes assets/src/HPD/HPD.md | 13 +++++++++++++ 6 files changed, 13 insertions(+) create mode 100644 assets/images/91.jpg create mode 100644 assets/images/92.jpg create mode 100644 assets/images/93.jpg create mode 100644 assets/images/94.jpg create mode 100644 assets/images/95.jpg create mode 100644 assets/src/HPD/HPD.md diff --git a/assets/images/91.jpg b/assets/images/91.jpg new file mode 100644 index 0000000000000000000000000000000000000000..befc3dd2cb95f938607bfc96e26fb3806c887b67 GIT binary patch literal 16435 zcmWk#cU03o7yflnT4vcTf)prwwJbqaDZ30=vIJzyR1icE#1_g1S!D{y9x?<4Wr&t3 z$dDzXB5xU1aDg}gMfv>l$34l(b8}9fdy|vg+?ZLIYMt^{1nmM>0YLn(u*0Q*sH@(W z{ZC`8gM2Sviu~8aoCydB^7RK4fq&)y$kgNG
![]()
k8h?6pbf&GX+uqG>U|@ijMWbeA)t8s=)KdorMw#VE<`R;L?A{s8I*mlSwbj#h zm~p C6F=D +!czY6QhDq}6o ^(z5}b>Zf!G>cp{oa z9boQs<9*%QcI4Waayv}95VttIFKy@l^ixB2s5}3r|4-u4J9w(x4q9%9+J;BB;i>p~ zbVxn=U$0y}x-AkO;)W*G5BxJTo+y{al$&AFc4%%p|Csrw`;VC&>I@z|gQvoG(C{5< z79O4TZv-kjgo^&xD@R4Q*~3Y012h^Ojwa&m f^d;1I&FYb05w8=jQv}%ze_p zK>5xRbAO+CynhTo20(BT^Ny;)BsibE`*=t7a4J%gRBql`Gn$D~3tk!TtbLk;G0G6J zxK}q(AnS0)ed1pIWHHY7S-HjihUr^cksnqj?l(TaO-N!BC3Q8;QP^Q9{tg2DIfxy0 zngv4#y{IKN-dwzvr{D~*^J lQgqSfc UK=S}C^YGcv#3zV9 z;J<7oerwc2_mf4hUlP_6)oQ?6wpD{0=H9g6`1l{62{Q;uZDue)bOfnE;o;tzV#w*t z6PK6n2&g-yCWJlx`3yz=ob-gkKJ^|TCbF`qPnr;BLEk7WixRC$yXV6*xC+=XQr<`+>rRw-o!s~%6(u}Yk5NO1atE8$3 zLO^zr5#P6yPg>RY&gKxf)ahgbdynlb05^X?NP@QW-!eW#blo;{Gtrub=!c&rWC2)$ zL<*<+)eXnpWJfai*^4Fsn};(;#?Ff})=0{lT|Vg-lB4GfhUEZCX0EPJQ69t}RSn@A zYs>>U73B&Gy?S@ggV+f*OJHxzriVLQh3p%mwW)=SoulTa$8G43tH_F2!7j2cqV65t zlS`Jd257UX7jtu2ou#LU3k1&sdn?_twfdW=%BwAhNFqG`1&U%6XIi&SY?D7;yiggf zy68*%{`v4uIjhEDuxGz;6mUW bgDEhX34?~*8v4$%8-QP*L}XS27VARTu@)^R>V z$e1v@A%qPZP =N< z=&m^!(j-j>pt1xX7x%dd!X0+dvpK-+Bm;+q9_+3FEIr*XAF~F1ntKHEt-EZZ=X2{G z4LJ~+%S|Ag(ud4@EerjiX6hY8c-mF$^Zb(?jt(r{S+lbOCN8_y4#UDKCaTY;Y{3L3 zyK$YwE0H-c8z4H-6IY{HKj>4ixS0_ww5T~`wQ6~)5-G(X)?+P?GxHXXPH)mzxy=}f z^BPH5VdDs~EQgLn-YC&R@_w)iG4fL(9eme*vgP)kW_E5ITV5dmvFaDSRz*H(+G`kQ zHhypGYX8yf%H{pvd4sP1Irc8yN!7F4;C5}X^8}}*^b>90xBD2oDSdI=dG37b*9)FM zlKqBx3tx1j-<&=c+b941cCX=jd*GXb IUkSv}jb$)NV^`51 zj^QBfJ`tl;L2*b1_s`J>>9V_#VRIFbCl8ITB~!0oopit+OII1%q~XrXuaB_G N&4Q&vB8&dsncJw;JgBhYD9^qU>HM7vR6@ z6=0O^iAcwd*qB_iNx0eMXUt^4!Nwh)ttXhm9<*Z_h VlkgUMMc`!L?!Zo`jVzxT;F<@F{=}6`$w*G?u{MPF4kHeJ|g!oYTAe!2UY8N-W4^oP~tD ziw~~z=TN%i!Kby{8?7Gyaz2pQKswtz@+}NP8R0pcLudoRTo=NNr)y%CEB4-l4Z_Oo z|B54^F+K*rpGjTcef9E%cy~GcNAQDqmKUdwO~T_c1Wq`K?xE;iKQR}tyCP;VNVRx6 z&w} 4ASW1JAFev!?A1T*rfINF^@6c>VfUHlA|u?*cXeonzU= zN@kfMo6=b_oUpzQ_{vg1e}O)OE(?he8RV1=i*3NH{0(zF;)zuv=UzA{kGGfyMJ>Yk z!y)}svt;R$BVw9fkeT=@G7Ot~oW;|tvfmmrDrr>1ahxe0cvsW0Ht}2O{X1I@&M8)g zX7Kxa92KAjUT&e;jXSIl80dI9jF$=oW}8S|SQBvhaJW!LRBUIL_V%I1 M(A zh{X#7o$yxl%RvCtl3fsm9xf{tizZWb2~2maUx3_0(!cpyR`=%C=;e z+zRisMFB}j)K9p*)DyHEofX-{pUKIbV8@_ufsRyTgW6nnGcLZpdTuUbsVXJQ !=IdI$V=hMjW=eb$ctA~m-DA|( zoj?8V`8{@Aq8WD7vE=xHxXeIS-e+lo9~)sMm6xDgu1%2RTx?gC;-zQwm$W9VrfZ2M zy7H9ccqOJS4Ng`A&`l$NkgGYUIF;L@j!5}@sNMeam!Oy~orQB*bstvs@KQe+6d5R~ zJ`h_Z;N*Ar`o*gk@WQr1r$w6hDs{|5TB5-#jv-w8Ga}M?FVP3|kK3^$ltd3l=z6rc z&db+N*w!tu&*Ao_z9*BoR)G{lwJ*@^55|`d-(9eOyUr~Xm2&=E=oq`ZQwVB%XRGHa zMa1jB9P^K0qSUvSlO7n4@4ZI4&C~7o5i2lBF68nG8LmDh4pq@J?rYl5-e{P8arZ#vo3GcSxHJMzToZvgroaUZ{mvh2d6^2jkS 9EqD&t(JNXMAQz{5CB z98ee;nh<3I>q5Ho$HMTEHgmypjfv{}VQLJ3Qg@20=)7xjD2)|T@Y43B1G1B JYdq_46nN9kcj{?&XAq==U4aviR 72*ZA7YF7UZ-IkS^3#W j7bc~-PMTY4^QY%UvyJCt#%`yDBaFYvT^#+PWG zKJUu_kS0b{=Gu+9)a93zj5JUcU?+h4OL){I^7zo=&2^eMB==Pw(1+u^f~ye)r(YGM zNOC{y(HF7P@Rvd=&teXjh1#4{`JH&dgiOdFH kBs~#97R0?m%nR`x zInz+7*G2FuA7zE?Pl?*43JF~)a#9O~f %T4`kk+O}oAZX#oQU>o;M$l9?~Py$(`_Y5+McK>E0&f?{}*%?te6Mc&>S77^~v zXBXQtAzzutBdaF%aKYB-!Y$#gQZ=H?gbRu_Jci_Oi+@gDCJ3qK1+Iz!oQQhSF%*4V z8T6|pk^pvCi?%V6i6SPw{GObtZ5_2vif|xp3>)>o2+@qW71|}okdy0tWq!Xm#aZo! zu(TmJ4B_qw^Y uTE?$mU*~` zEeGK;wy#S^)boDTfEZFa?EW3ED_^N0V;YFy@0rl3i(zVzB0D640G{VT7?Z%9wsN<8 z5>;f+7{JX*`VxqPNiuGSQMg?Aj=N&DOf8rN^q3%GLuqds*RZqp0Em29kS za$5sG1bKyKFAGM3=fS1cyoAC6%kdzd k(cw_}8%gslM!2F#9Q4aJ8*#S Z zQ&y?%bSA{Kfl5ACZB@RjnujyKSq0-PzJ7fi&5?oUuxm9vZDr?{^4s2;%^&G^9}r!a zl|%@d-)}t100I&fqMUMsa9IkDu-Lt}n5PvzLG22plp7$5@l$dSPc 6Wm*c%%Kl+-*< zxh*VP`>u~*tlc7U++juWH7z-N$!ZK=Qbdzq255T%yB4QZW05t^Vwc|Z>SZF;8bq>( zxWS!jXG?fpZiNowvzN~WF6Vo-E1XV9b@-NwWM4tHL_Az-29Q{wI$h}O*G8$`Q$UQy z*ko@R&y9vmn`6YSk30yk*Zuc2)w p`oweVX(@{uqyd-mEc^7d@VJ_k~iV0z$xQ1!wff8Q0WS|aCi*KN`IW+ zlCF_fhT9cHOW?KqyWpL1sf_}?Iz62&ym#fjzk!8+_}V+YL#jAO6wA0E=EQU8zTb2P z(hl8&IG_W^7m#KMF(e(9cPQA2=x%=M@b)lV87wAxwf#gak5ep66LuXjh)}PJI7rEK z6i*mgJZ+E2-P;AVJyWUMs6OIhg=2a%Gm!G(T7Cn+k7RPnJ=_@?;Lon2ZKn_m0NN f$02yppwKO)Z( z6P~&B0g|`2DExzaeII-9OHh1P&frry@7hSY(?n!eO>47pD7=@G>@QB9gMmR;s~2Z# zgxMUDzwSSQ5900hAZIJCma`E2>;p!sdwH(Svvf`xx^=M7VK~cy-EW#B*zmU0g|3w{ zluj+*nx9IKL)|qPmJTZ}>gD+@;W3Q#Gg7{Ff?G6n;=ISXVoQMrc>xwA7|Q@df3{$K zGMhpH|C(u7k*8Xcs#SN;`S|vrs3|LdQ@ t;iW*BOZ|>lIe_ZmNUOh59`v#^uM}ZC4^ pedi^8zN%ubsh4FZiIZX`S;&(v^$D8}2=B~0 zMac6I1hW54_~b^)<;(>Ky(ps8;Vy^`^Qoi7=zd{nzdejC+)P}ML!#p}j12i^^QOOj z=JcrQLj|6$$p>t2q*9%em60;ndBDzlfE!p;P(FtWy{?@(4|=Y5@JxeCQzEGx9V%Wp zaWmz{3m$(4WHK3f ^gAp@REfc|hw-159? zW0InIU!uSm=#jFyDD^l@ccZ0RDrazWWhmFmK+~%sK z4us3=NUr~zrU1e=DC=~KdxlNtkB cLfk&0uVd?*IwbrmqbCJnm!5uV16u+izMi?=3r2PJpi3-1{d%g zDd8c5%tgh~vv7OgF%a8H(?mw?`7y9Qmuz@d1UU)D#y{>mY`RLx%H#bz**Gght~N5s zSUk8SIc9EdczQRFKp)TZkLUP(A!P&yNJ%TnE?5{7AzGW+Z_|vXXERBB!cwFw&F+2T zEKLznrqfUxsFpRo<*JRxJx+J4Q7t-4Iv8-1(6l8Og7V}9P7$nJz6I0)g1YjvxByd- ze!N#v9AG$8p1p0Vx`Fw2Yz|@ wLnkRc?>Gw<-cVv>Q?CGqx)HiWP+!{x-M!c8Vyfu4U*GVnj z>)OK{+lh?!RyAzwCJV%tzeDb)C&ZTiFy7HzHC^4Us$l^}pr_6$lgONKQ#bDMzM8Hb z>0@ t57G%%&8guuU5!Q#YqT9DE z@Ta@YoK<+~$w=W@bv%pk?%)U&mPRnLk9eIfH-nM7jFHV|pA|! PK z#=#4~=@_cn(WFzmZu;9$?|UcBy*O`k5P-JKqX@&^%sLJFo3qLOF4Z?gV7=|vVxP&X zQZ^$3H6uW2Y3(@$M2)(ht@~+@BiJ+80Jtq59VB$IP`)~z3>T-k#bLE2gDkmR?~5I? zd9`fNE(q>!ufvDSmd$fDziyfN?0E-@r9kP)Fk`SzFoxBoc%rlP6E|QkZAnixBNKG0 z!6(HTRu5V4qTPqM#eSX(W4}szmS5=zw-7s;i~}$PKEUGwpFHQ~IzegI=Jwyyi(2px z%vTy *a?$YEycyid%lGYrr8=V!}t}Y8o z@u2ctQ|T8m4>GwlrGs4(38$PP)zr_Rs=^)@yc#nW)M5F~>7)-N0>ZeKW3#EJ01^iZ zA~F(NuY2(dymos5E~y>ifq7mmef2)(l+hiU^fj9J#EM8286D4d=4uk~hbV~XF-((Z zjWLGkvQQKX{i4Beqx8Q Z{gp4>mWiG*vkC5-4 z+mbiS>Rftw0@;pL80nC59Lp (ZGaCu&TCP85yq~m4 z?GU$XXErY;D3mXQ16U9ssuye&KL}TJ!Ai-^vYLLX_|x>gzrgz#EA}TaojV|$9FM_d zOV0|-OH|*NW2VQ|_=(Y4)0IX=^RHXql2#NbvaAgYNv`(EyoVL6uW5>uvjq*+|BHRd zPY<%9c5eb&g?BIL4JG CQM64C 1v!=K71uW3N$9XYn|liE2g32{obTmC*EF zI~I9Kzf#**7G(>aL9lvx0K2D7$U5Q-gR+6ZORmo+#_pZFFe)l5wN>j;vvAF9O>{+= z{%k3qY$eCzf(_{CniOo(V0>tFP_w6xs*~?HR*!$;1yd(t15{tVwQBnQdd8WjJ_B@w z^_^&uWPf)_oQImD_w?NT^I+hv1J*a{g7##U945*CN5HJeo!nxVrEqH>PN$EF&Ux{s zCn@!l1}^8uD1i%)gQ7mIKYFII*B;(#{|x8%pIm)KbQo6^0)+fkKS;@Xeu~SS0AjkM6N{hluSxETUw0f8b+mAfbJ-+-1?r>rme-EQaBk9jjrNN3`@eeHs zSr+Mh?xkbuIk{#j2gBlSvFRGd(S;zcs>%>UlUpO7eIc9cU?q ba zmgE+nv (p#aAsvmAhiOqdKOA_Vu2W5RN#o#{;tu`EXS{+sh zz4=p19SAP#r-^m^XSwwHmnN2?NT@zmkthW@M*1FQ%$jfeGOiZ%DMQRB`ld#-U;aTQ z`AzWq=zp8#-tT{IAt+(H(t9ZloD)M`OC> _A_Bon^ zc}HlmI})2vfLCh0=!f4J06Q>5C}aM`$QfXvgTXhG{mur%AGUMZzJ~3@^ynh0Qg!{) z!2t!AKf7tXmrqsU%v|h)(v^AdEx;+J8{qtmdmKVdzqyUM(^W?XzDO a0BG=^j(fk=3>It$yJ`)@jE{+`Yk3flGIv{(i-u&%;v>j_M!~tLH7x` zu3-bRu^a 7;(_sO8IW0G(`Y zzMJr?KyCJv(H;Us&w=*vCk{W53Q7YlR4AwdNUdr@TxQj98l*}sR+e1YRE>l~Aj>^j zWN5$6>5qIRjX`B!bPRrp8zd}rJoOp)UfUG?OTI!iU2SdPs%83X?FyQ0Yi?e(fNJI0 z4+B^5oYFOH726HRCl gjDl~AR%-JG@) z+QL5Uzg!y3##>ME>NuU~VIvWe4?rlg-f!V_amA#;k7Bdi9ACb2o#ip$dZY8QUrBpi z{-GS=Tc_xZIC7z+33U01o*$}Z{xl1@VT3_L##GFdV8P;WNciwL$wc)R=!?QwZmG&F zx;DpkOsK9c(9KMhcv@xz&hbXv*HvWHme<_8Gvl)1MtJ6*K{Lk*<18 HdY~^0$zJ zD#$1`>k-W{t%_U?`oQt!*?@3OIv1~E4<2Oxk;>aq^Z>)r`=%`$+uH{p_NX<>fet5G z)zMr_Kp2YihKP-vFc@D|DN!Jrcwes9Y O-Pf>1 z?B98gI!%W<2-D_wZ`huYZ!X56gx^o3sh4HsNa!>fkF9@Gf-`DQJtKlBTK}A0_#36Y zG{b2-OT3Qo;!Pp61i~0HI8U`2owLj{%y91-E=U{(eU+0QM8KM`3F0FpaS_MzwAPPW z$60CIYQ7MWSz{D@`f4rjFWY>tH{I@30@K)H{g#kpL~8@(R_eI!@#!huJdO)NC$oEK zFX}SDxiyJBr&F{=m%=GdYxMV1a1;RQjsQW&`MaB+|5E1|9_Rl>6R6PQXdQVH^N7?X ztl#)uV_&`nhodBd!Mp^+E%wPf5!$!Eg2bPR8O38k<3ld#3b|heT^8_~TU0GfOJklO zDkBLVLuNI9ED$ya_Oe5H%%QH6!DKQkoNH>Z=-i1}t~VXFt$H0bm6=U^cl&eCdG|C1 zbL d1^rl<0@Kx`kIp(ru7W+`7;H>k8?##6ZEy*Z!BV>|JgynDymNb+vi z?HTf|ift7kxb=-FV53PWoo@OoRtFT*KZj=P=}a3+B)yi^f?~=<<;k!tQlznrNiwY} z0=rv_ZYs_88nIk$zWXtqs$D|X&`MInGief zAbQBEG5sse#mzD0s#4jIE%5@RqQZ3@)RxIEqw*a8nLAMQ0KbKM9DBLds5yqbrv>4` zkiC&)Zxpw!UJ@v+?|he@&`M!hQ(xw?*)m=h-fXf?p`Ls6SWFgj12YiSs&l1$*vD=D zp7ztf+6CYazK^d(Aw59driRLpE# f9jA1jdlv6l nA4+oBLAaz|Hm1G)B*T!1c z!oYILhQ kUP9p^Unzf>DcnA=};B`?UG z7)*S2Icqr!wOlnIT2JmXk?3weO!W?jYBZVF&@-p;Hz}kvZItkvsk|ZMtQm^J8_+ rdCdlYvF Tdvgg<(st zH9Ql3FM;109gX7UGjgB3h1(zpm+yP6L_o8*iN~Kmi!SgRY|LUx% y=6Zo9VVkr(HiwvC*w5wT@GQT4L2X=LA!@qg7XKb}~28S8IfGd6!!(a_9z z$v4X_Kvpw<&0Eh_BMK}bD3@Ol3O~`x>12rq&wySNj29&Tn)&%&<)H$0;$xn zzk=whS?cS$a55R4+llH|Pw)b)Z@q`e_V@=RVM;c3$W)A1l068_{Klg>+E(>vy1Hu9 zGqdWFzn5t}dz<(RDg&AhT3auHYKLXS$MBhNv6h`ALr}LN;eeGd8EzX?2m{krF??y} zX5a4Rj?I_D!Pv#Xrfm+9KwzF)CEeAStKiRL^qw|Smf~5MKCyLsc2Zp{sW#!^4S M2=JhXp`D@7bwJ1<%?|m%)f`zrfn4QR|=_Nt=i@QDi zd+<_?4#*?$r91S`cWc9DQ_9{ayz3>%{hXV;r198-XSVUx=f_Z9$}Nc5?$V{s8u*Q{ zkJ3-6qx7@r{~m3I&42m0_V&`pg!$V_mHsR<8ZMA&CDWgMnHu5K?-j?JL)rcO`;Gdf z?pdqh^NXjh4_H`?@911=xHXf#QeT6-JiK;Z@@X5fcJnOSEBh-CI2UrW;%WH@ofoZ{ zAHO+lY=s}%$mocaAD)vt)<_6W2@3oL+ZE}AEGFOmkrocy^-29=m#yrH={2EfTKtrD z@c5upA9Cplv*yO`B0?qds(w!Wmz!BjDk3k_&czSUoP7p6qM)6fyi3tT%+F!pI=0*# z<2gQEPoC!DhzgGuuY1b@4yOvpMD&rP9vinfUwHFX`3*7j3UeyVt?_GoOj3_1#~+z76R#qy<+7A0&W(!=K{J}q+?S@9?LO#sgb(~5C z9XM*>`0#EEAX|vG3j|EwVdvk?rN4d@QXb8s$M^WV;;Hw7VoG9-w`Fc9mXZRv|NBm< zb`p5LdB)=h``vF|lUI)WJ0H6a^rt=1r&vU8n|k{``H@!Cx;7TQ!UOyZY4LagJSD&n z##}Gv$)(#{%}WLId;@bKJJ-JbOerXuwk8RyEPtd7B)pDZovDbqEgl0_z?^?7t82}7 zqcC$njKp=)#H@VmL&tAba}wp6?~6T%Uays(ic<8g7?la5p;?W69o+tG-HlTl7_9p* zLD%_HrdOAK6!9K-20s5O>Z0=Xn_y&E#7s4LfF$twA~$!b+p=EhcYNbuUvZ`l>5}im z#gYFGvYba7q$AGlKR$S+@&MOTE~3IGr~BXO$j1fu;BzfMnizBY2h)w5C6IVbHtB^7 z+e!W~Uw^6P=V|j)fn)uOYZd>k>#F}cdlG<@q@Q}dcD;gF=3GA2EQ xb*8Gfw67CYEoJK>(|y$0i>j?l+O9A&JmHBIZS^^g$xBIM#7G znEfPMq1K)l6@(xI7 z>YM9}922)#Bg5Qoaz$|+$?R5>VZ0vpSusH3i@Ok<^OU8U0=bgsG#L1NxK_@J5ad7j zK&~gVz%U)|NDz~ZK9 SKPCpGakX~LWDd%tF>Gg*( e$DyPDWMYWEJ(SI2rmZItl)GYvfH@g0vPvQ+;er*F+Zak4Hz54 zTB89YwX9QlRy^opGIzL!7s35G=k{D!Zl~;R<2!~wF;>qC1$=LENQD!y;<=8w$I)B| z3<=0*@fZRak{eZOiwL6KGR@6WFC-U=HmqEUomi|+jvvGVF*XuC0Qiak;uV++<+ueV zS<|`02*BC*LHfPBs6g+&S&yI`Uff%sid6)aorfP63})llEi+TrV$Ax&wz>*|TmsVW zB0Gp)mdRgOFLIj`OScDYD2+Xba2Ns;#>uDY+dc$CMoUwhT=>5XD&6;BFsD^=unnUZ zM?Rm&KMR1Do?R o*ts zc9nZlB^jJqflZTKTF&n?xZZu31xwp6YAiYYQpkHHpuHOyDE~pU_ADoj1#85*C^ZMx zA(CCeY6RG_G5Gt~7uQGw^}$!%+GNF?xQ}`~2KtUac#Ikwqz;T&rdXORj&wh0xqdkL zFUA|wDGE;KWGpM6^)2lJpxobAe?xM|p%)*)IBs4%$0pUa(Jv(xuJyg^tOglzrO>Q= zlDEzAb#PC=f{dq|l=^|2aa&;yWKPVJoSDbTH4=X!_&jjz6!F=~H99JhTkc%=$K!z) z#04#ky1!IRX2pd7M>5*HUio;^bHgoC?tSYdk+sBd&Gf8c`mT{Y>~v*vpOfIz5-xaQ zQpMw{^$u1#r}TEnIjAka)F!Co%8`Og<(4|xSd<(fTQ| I(u64OMUW9u#L$nj|<79U=d(O!7=##A57ev!#AB5kyWQ66E zdNpSLSXqg+wdxDkp`iKePLJTwI%&&h P&m z9if$;;W3j0VP2!a%{tf3G>h7ih(dwh1>75=%(=1rl250lrNzHldOIz|f-$@{TnxaD z6d5w3xO}Wt_Y_;XkpJ$HkK1CY+1){9gI~ASHNS0={^B`I#kVJ3xBB!_Hiz99rzLQ0 zOf7j{jFhTXUBsBrNQ~f3W3BaR&_<9(@VtYx@-s7&Y2$pB&^#41$@1*!O)c|BS%6sI zn{f3qAxU0;>Frh fHrnJ!1DN65!!NKvK$nYRl69ov#59-*tm z_V%7#;;4Gm%%q(;UgiL~lX=b+*5`*>z>lrC$b2wkvYpi-$wL!vcE{1S<-W1Psaoyw zvU#kS#WhR)G9Ll{v!my?eAAtTeOf-#m9-Kg0B65|>OG4hv%$pIcgmcE*Uo-cuN(Ok z*_8&AjV6}|Od)Iy)lxxLgj0FTXD+Bjy`I)>#dhjV`h=x95Cg{VaIUfXv;EM0YFKo7 zD{lJWC)}Y4zjK=Jt@O?(aN&DL*%Ge_vre!1^3twp(iQx4(6>WNWjz@%vdS&&TNEHT zG-7ummXMFZ6~5w#NP1H{5kT}2S#ojCXv`}m!U#DKEs{Q$tvl#IJg@SO@vS#=9{c=< zR@GV<3ozp;)TypV7nFRCEo`QY?RYk)`r1gqr@h WK( zNH2PoUS_HAvw&i|P|SZdY^7(qUGo&j1{x-^?TznjU#c91^44 uZtE3ktj`G<55H9`-@}HS$ib(tmy?sUY{* z?~zb2g`${g%!5pu3t&MXLrZ>i{d$wpgR2eZ4bpb_MT`dJ{vm_m&*c@NLqvC6k+vdC z9kn_R{<7W>+wVo%cG~Z|b&I^Eh2c2SEDT2RB5wTPnvoD+z7W*w*Xc3|Wc-!3`1Up@ z<+ZDXn{HuykYnk}hiD1N^YnC0lRS!Dp{&N`mvzQIN>Im}q1@?Y?(OJDjdJ*Umj%I0 zv8QA{K{iw5T2y DiOD}zx@9mtWuxq_WZXCsjn=cff zhoL<=!2J!3*NVZX iiLW*Ly@OO702dNJ* zcz2=MscpCHDdx#ln|5zm?G#ro6OOCE;ZrEVJErdpHU*zi&tg$t4sD6>J2ZyZCld zr5705C73@b5!-$xahU^{9k8Bb;aVqQtx%uA1cGtptE|kXCW)-g)o-&A0EkgRZUlMD zTm8;H&>gw=&M}6^%X54&5=S;vEcWbX8K`i7NT7yxjRieQy|GpdnSBG&E!L5G4#wGT zBl&xb>1ni;Zl%0;Rwj|Qm_TitHASH)7f+C8B|lI8?ExM=bd_}pE1XD8?3nEA;rfZH zawI&xX?&a}TfN9VG&8a}8L8;KJAm3;EaZx2aHrcRY(c}N*bR(VBAV?561bnxFu@Te z(|>HY-?G_C32tVw=nsr?ekfd`o1Hvsa`fY6JHcVXbz!^Hl2?EQ``JYHihxaTas-yg z>|OecS1dYV4iBfjh_0#D$=-NT^e3wYo$Z2rJ0EGm^%r%#*tZr39;~kgtWuEq!s4s~ zObpu}yL?3I^KHB0uW2Wk^Unwwm$3q;zdpB~uop>?K)J39RPZb4*PxmUcr@%sULia@ z)?8z21d2RX!X!3P8^&8c3jPKAM?N?FH#DCW@m)9HQkT&Ba#G{sy6R;K{XBaSCY##8 zy1M5v1M~smL#LWf N?+?fD==oVc|2vK z$L$P?bS3Xx4FT*$0Qbzk@!+7J_mJEOl2B- 3-8HM4A|W#oaOz!vOt(1%1`pUY z0`D3f&$<5+CHsw3y<1@or=2_B5nv`*uP5N|dpZB|CliA8g-gVDsu6y+bbrlk31yQP z0rSS4C_>!q7_Itg0Q}~{+YCpq+%!tDaZmawQo1f|BXkeG@yg3le?h}y#jm=|Pqh|x zmc;-`z;cjl^57xf#zsoTpm>pni3>*M;yZ8OBiAQjb9GpXjLo$FjmQ3q*4W%UJF#hB zO1qS+`CGGj49T|XU$*++syb{yfvN{Q`$2;`nD6fPmfN!h>^*zb+J?#capSh(43~zG z-)sX?xz Hx#${c|5V|%27Bqv{n@QTaFe1>v&>jEjl7jL zCaH6|$xb;yjxDfco?%C1l0Vd1RZ4k8r %v-*Q6R&@0 zlCQrmESHnDeD7Yc85fZIK`1{Rd4J4D&5&vmO3e-*iXIup5lo?3$M5v*lOKX5elFV! zf3DRHJHNp};6AHRU1_aUKPs7Q;`r9qb@W5GQ;j7tk-aO#lK`Hv7~!@8-?-eIOe0Gg zCmN`?qU%6s)jt{0DF1~zpsk1&n!)F`B(U)W6h8sg^wF4R2 Z_6}Dt>%{{Dt4 mjDJ<7mlCiie`)rCHOseQT@IuB^Rnl8>NS|j2I1#)QA#W?$$g1*DEu>v74b7k zaSKc~m;#_w`d-R2G&%ft8wD+U+ TMUwv+1kyrb4^ISjmCV+pVr2 ~|r;$DgnfNO0rsegaw8_KZu|Gwh?gqoou{6nXC%b^J6`@YqQ1 zOgP+%jD=IixX5T!xqy?5wMfUD`$gP78RO83H4gh{ZI*JwS!Fh{ew9%$LG^u-#+eRo zjGEpQ1=Mvsh`zsB^Gmk=%h^wR0c|ijW7+9}>vUW^(9|?&unVblon)OQ537cv%+1FD z3LHiMdoNPgYij9QyE}M;d$~>JO*>o!(A(M-(b|FR;ti-EWF2Knx&uD-zTV{(@kAIR z20RVFU{MS*xq2?gt2Ud=W@<(tW|9H-_K1Kf6UHl!k}mdT%5TrCn69X$6x?ENlyC{< zZ6AB8mqhQpT*JLN&P2j8@uuMY VL`mIc(2pYNiKhcFViw&@a+RjTS>|?ZNVQ9X}#&{FG{7w{88|+1H?WQPXq9x1sNH zCGMxabiUh=eB*;JR-MtLJ*?
;&ql=4Wf io>f+!IP4lhdTaBi{$z&g-h#*Ry1P zQspZZ40gI8A-F2x^ir+TDFR|WcHqqziP+@_05Sa5=%Vz)dx2#*Nvf2K<9ry CgO33K literal 0 HcmV?d00001 diff --git a/assets/images/92.jpg b/assets/images/92.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d441731d9a5bfeeded910458083d772229e28111 GIT binary patch literal 27482 zcmWhz2UHVH7u_VJ&_gFQL+=8D2nZN@#}GhiBE6$j1wjqH1*CTnkY1!CRTCg|F(4oy zBHe}ttblyx=ijqu&wc0JGxOe@+1c6q#>m7_?UL&skP7H700{r%v%c*WcGoq;Qv+?` z>mG7D^vpzG_44v{_XOmCv+MuNBpQwO=g%K1mAbdSS5s5Ew7wJ?6_s0B8tNaqG`mD5 z&ql?^`}@1c#}mk8a!X51E+KblXsElp8|UEQ?%+ @p-?iCKo|-oAPGuJO2R0VE((W4;mFibvMzx@4JGIjy2
)kDTdq`7M)cPLI zfr?v?QaS_cN;p(aROucLMa7jyDV>2*C6q%=)Oribfr|P+rv %L0~e}+D)lFXB5|Qe6xBf&>3>GjrIzYa9gtKMl3I$SI)qYDq14h)a_LYh z;p|x&>aL44#m$ydXZ;DYxuK>gBm#+~uItw94Uzqa$W$a*dWhVDB=`>zs7QkJ5TWJF zdv=G`?3I#Dhsd)?GGd5)ma6FxVHQb13=s%tz0_zp;Han?YH3Xr!JkI`6Gf zmqz;IblnkT|FiCJ)b$!NRTt-9LzbqJTXbioaRh%VfvQW8rV`Fff7Glrs-~1$ =1?4|`HU!8s zKt-aY<5AMHsQ(jMP%TItE{ckqrJ`o3gjupFb=DL|p4G*f){qfYGC_A1fg_lnd2|T~ zD&fpDMa?44!kDcw#nm8iHKwQ<1j>QX;xN01Go8iFIv~%$EE0!^cOdNH5VJUf1M&MhQoaYH6XOsQ=Fysf0vINBI+Ub 7PC7S##8NDuM=p5CX<+6{I)> zhm6xiTjfX+Qq;86xV?&;#;feNHqriM{1I9&Rlua9dNN1Ky3uK}qh`85(S54aq_g&A ziCXCAwaL!9SC27q41%Uz^>a^*b7XEzbv3-M!#*)BGwt>xGXmf3-c5BkT`vVAiy8lR zH?MU1C_7{7SYvzXsTWEwXZAdOHxR+hjZ|_401&6X0K!bQb#o*+g&l?!zPioe4|TyF z@VrZ45v8-4DDLh2@JdNgbhMER4`zmTsP0in&gG0HI@9Ip?TRC0Sm2L^7Dt!c?};V~ zhLU)>{uhx#9vgC)_+$jZ*X%LNoT~lW!%iNj0Ikb+H&$C&X{Q$0LEZ1W&*JX-+}eO5 z-pU2&A$zp&u(#YuMG6RfG`|3qLf--NRb$}XV!i1Qz^P(|jzgak1ET{?*vIgi*5N@g zqB~JtkPlB(7t8k%Nf8QJ#)D)VlZ_yX!w3=x9v@1CT++rr$)Jmm6IDMIz+f`r$Yb|M z8tchB>AI`EPcpFIeqiXV4PAEg?7SLk@)=5%!Eo`2tPPL@eWnpSSDF|DUQUt7ygyIQ z7xuxEMqZ{VZPDvMqEaV>!SW$eA1fXXZwsez4l0A7NXtx}Cuz?YKdN&ihq6Go6hGHL zp$24cM5v*RDS28X*etQ7GPb<2E`F9R{Zi5vuu-xOG zH;CT^TOGI{rwc1!4aQ>V7?1~7Ne8txT%^X7p8T$pu`e9}Vp-P8y_#!3R2NW{$fV3| z{>{23&ms16c+Vtl)*m9@7Gn21^T-uc^52LQ-JSqaM}y4{LvmV?56CYuHM-<5uUa4u zJY&7&amP5|NA3@dJmY@29mg8&Iz}ky&Mxna3QwRV$U0kZ8xp^wdiSj(Ie%PT_Wk9s zwXipu>(7yk_pL^mLw-C{=VoVmI|Jn E6Tw@;GC3oAiz(OI<3IkvgoXF&gb^`7oiEW z>Yr4ik@k~M?yCp1vfUV%uM&)=j6$X5kTmuj?pJ`zxuckI?Vwm|3ZIRFI!WRZBfZm1 zt@n$hMb?`?-#T5pmzejmcRG%45c^8asF?lMK|D_^A0lP!#e@`w&UM8cth=XJU2`?) zc}$7U(PGPQCKlgnUji%5T1tCj=TVpnhecGXwd%&gc+|fi8FvwNnqLLRU1-7QNT1y$ zkcN*^OB~k^yK*$?5;J|ce%#pD?<|sKP>iOtulk#4$9IPea~fANBF0PLz4{*A@Zw7b zofsA}W6^;zdfUu(-+Qub=c25#-Wwd*wq_s5KAeEK>Qdlc3GxBTzpXv=MS>bbOBLik zAdNtDLFitcS~uwSjz|uB$%f!)Xqj{dP*srBhBwRSoAbbNR>(7YV7y-z5!K}mSb|Kj zft_npI};pjurU0^W)x^bW25v|Dgdcw@UB;_s5eyd3y{$D*05IZxr0l!qRwzuqq|}v zi*w4vPmr3wgJ>)fb}t``vfHF0#Oq-JL`DH#9VMPAN}>j5rGW8ARgn|%89SR2^hmuS z$CMR|c`gL`E4FZS2ODS*R `4mmK+W2)@L%22`Qi3d_6XuzP>vMXF@WkN&zLu~US-GvpDShyEqI1ydpQZPLr5 zxdMI7S$8yd+Xgo!wlr#I058!M9asyjeFSdsd9g(KU<5oxydYz4#i7f3U=g(>#J;S& zQb>VRdJ)ETqlPsi`j*% XKDLIgR22Lmy0$cB-i%99 zEPPn=sq3t;7HT_T+(EsxTh}<>L6<CaPs z|Hi8)##pYd_x!eqwQwO0kIN?(@$_gqC;?%n6-DT3WqZrFcKz=upRAL4(VK%N)y zq+A?$9O3_Ta3=rvc>r9mtnTIgk+k$#Ib32-FKmDb;Gpz8RM@%YZ}L`gkp}yNP_n`6 zsYa=CDD$82Ip>XG)o)9PUJBpQl8gljG-Qim9nQOaVj3G;{6J*ruqx*YVk}h!PK<~n zLz(b-0r^wE@HlXa;mPJ(X|Ua|E{ptNN|&Un+qvy9Q|E7v_h9&w0cSy`6! rZ4>7TakyFMA^h=F`(t;(77 z_Yc_CUq61pF>fvAcHL4 @a!qkMD(JK#! zIA1+7sAQf~{rYL;iUhKY@s`(Vs>){Y;MJiMY-^ld*qz90e5ZeS7u*%?+8=T(s31KQ z-;ZC|zl7OfSHfBYqAu G-xh+)~o!iPZTuUuk#K&MnI5jJ|w=J}pJBG~=C ztKus<^o9t4UPgax16_aS59KjJ?!;*>$CX%n5`}$L^%b9QBdbY @)i9{%2dKHzPndo|I*+N~?g8`9DK{hm)8BjJk;^L=hr@<3C+|}L3 zx%9l05j^H_VY4|mL&HNq(F5_IIQ;|4IV47;=fllkKhEXSu5 ;Gb>*fI{d&}5T06JZ&tNzmcE(to+OBe#G=&_S-$fmg^Kf|K@3O` zJ6?R?D9y=J+?ESuQv;9-Z5aTENhycw8p33R?KHxn_+t9&Li@8&->9Xx{b4ojF>N?X zFYsjj ;B~xR>{RotaIx_(1Xnn}iG3!U=WZk0|xFuXt@8$d7Qp9#t$7 z$_&z9s{ta`04yc5I4`r xCkqa$=pjT7Q6=**G+Oz-4@stS+QSOtV!ODJ`1hBJh{c3 zrycoxE7|gqIp)6k?-xM4M2nh`g6a+sui{+;dYQ5dpq-xuJ9IMmm4aFkS)a%pdz5T* zhfH~ulrpq1`;&CVR$(-K;l)1r#zR(742rESrc~7Cp{qD^?ES2Gfu?* _zm@ z4w@x96)icc8@Xbu1-Q{%f64uR8e7tNXs}2^az#=i+?B6c|6`-Ve z13hIPX5{5iUao;>k2CUpnRK+mvE27{lme63w~yIRD*Po5la+)eJ??-YK>#lv%Skcf zIS^Xp;%0RLv0*e8Kk-meV#QhBSSvGOJ3jBEOP`n3zk+$eNT+w-k%x76>Ltq#x#D@P zTxg+{ (7Rdu`~u{UGA zkozG3a|OU3bdJG=_1r2~zGUMBG%$fD!e7IFKf8jtvgNY1J@f?6PXV?`d?6p}0~pwG zrtF_CwwQ%RMFxk=UuYVRS2-2|q<;7@0IbYGEQGk)t8=Lw#CX$AVYW()>vd$WjX$2D z8J<-!k?8DM>jv&oQ0}q=&|q#1Dqb`cO==OhdPv=C--o8$E!55HQuzHTVW+aq()B5G zM=ks?!1(h`MaKuaG>!;wI)kqrA^JOaj |?r>V|fz`(ncPL_?KQ#&aBl4DmJy$bSgxFMM* z&e>SmLGp@^d(B)7KLZ{R#DH=SdBF#OU?xyRKG$(Eu*y~68-(^&t+sFCk{oEW@iw|4 z%<>!4^&HN6{jEZV%$ ?#&IQG^yL$mffHi-8X{_Tz?AZLaQ5vW4GqvR8L#{^+r~2~ zO@my0=_j }FU*1M;p~@#|WGQ-+Xz&)v z5}64X!OQ7(n?7FF(q%wr%6dM(<5g+KKZjFL_Peux?LPDg5V6rlaAhSL1^iwGLi#;L zZndKKa(AK338#`@6zy7zs?JyZktaw=)CQ7IRfu7)sGTtpK=!L;<32x0FIV z?a;4uq^XnCZLx!b>3LR!s$HpVQPK$D KTXoD~B&U4twz*c{c@lR*43zijow*3J(Fz1ioauthdsMFqt(&Jfm zWhWmV&Q{LYHRweCt5tJTvbfXZ`NM2Lv{KPx>3l2QGp>;IW6l012cEAnaOlBYQhSHl zyQq#s;Isl6QPKWmh8N@32Wf|&%5ltX0dJ$jKSC %7kWaGnms^5E_()p{MnO?i$)XTIxoo&j9nI)DbYxjSt9j#~T *~0 zo{&dY+&B9kT?28Q=z7hoQO)^o3%Gk@4a;+}OI04+vzW4nzjYLBe>m&K43xHvJ$iY8 zmGXwZecdD&^zI$yK{`*1VDDjw=TVT!t5EfC9q-ONfi^J!o8{{izlqO>JU`pn?7p(G z=yCr{AGxR_L~7v%gBmi1g!U}hLsEs5w3RH#SzMvZjBEzZS#w;&3R7Pv;8`0CptW3n zjLcIuyG?+{x!HhW%HZ%q2(zb>P!GGXrsKD#+4pqYTvqV9KQcSSd2cxlfU^`DFgD)O8jG|I;9 zSZ4|RibgVGHd#putiTrirfcMVw(Roi(vc {-{LzY zBS4SmQy-OYxX*pD>W)s$8k?)0n9!7m<`TlyuilCG;&SBVw}oq*J%HeqfzR*QT!k8) zx0ojIe*?DlHQq&{KaN=6(qIu 1xOi(ew4E@q<=^`}Yr@k-qZZ cv;q`rV==!1qr5S!DZPSP(hO?*)7y* zjX^=o# =RW+3>}Wczcw1i&7|!cS>+ z$^|vfXo-Ty2oos!N1W6_%@DzT$ebvH>x*0GzHll L~gBQT=jsYww|>T2f}}_%h;o5&0;Mn%>`aG&^vdaqcAjMaE;h zlLo|by88*v1~|o|nw5Cxzq9hU!C_c9ENFVJfqJFAS9d-svz>CC#LDnoZ{(g0%eGFl zzPN^nIJV=I6)lb%PMd!rzVJwVs8{^2i#R%j1x2d>MFmL-fB#SH`+w)=uZpo_hYYY# z1F6^q$q07QI|;0T7e$rXuOk7w=vdmHSCUHewCMS3Krm*52suRvMBoodaY1z4=WTka zn|k!DVulfDOJTzdR<& 2TmO90X_iE}DdN9Z9j3R9&XHjw0AZ7=Q@~IyUhEq)QNBtN`N~U7>?q zL~f!}fB4sRa>;e7^Yi=15*`Dp`)h5C>`cF-3IzHFxM2LeL9T&Y{ZK`p {UNAnR5aSd2&Sl0KXWSfySTOS0PEkLG zRL&eTbMf7fqWiok%-ehZ$If)_yG<1S67E!;VN1I7%GH0T@${U_FOpuZ|95))_0v_g zOKy5)%Z lnIqk<1DrT&D4J>RQP5_%j$i+JyvXiESIM%?|OC`l8x(UIG!YI?IsP z8{;cp=vUPs!QN-@LgsSR^o^xxzplV*3cU4>>t#U-gMckvCOnuW8GRn0RLYr%7kIEY z*Y6iM$1*TUxm=K7tCF2ypu+$Y;xDa;c!}|9bI^kcePxS_Zw9~MTbJXNE!BUWe89K# zpeoJ!XLW#Pus$9m56mB10bw*cyN?s S>Yf`gq#teucRPWounUcd> zU8%*_Oh#`F0+Uwsj2GJ(Yy|C|y0Dfvt;buxT)1vBazV1?{_)AhiKrjsrWhIMqQ5+$ z( s Rm z+Zw$M`vB;-=L7im*ZJyRNl-|{*5Hv4&q6PQqT?c1gm>dx*Z%X1APAkBHvot&-+9_l zwx{WJhktrR@xm4O@p$N1t=yxw#q!rvq<`Xlp wad6>f*!cU|<@>gmrYOAj|sD;t7w zgSvJU@7sTIBnC;T)s{6UJZE29fLXU*M;*q8A38J*8x{EGQEakT1P>P2^FxX5fj$uT zksxwCNeTcrt~&k+>K$WUs) S|KV2U^B$Df{+)vZOYb a0Z%itM{F4Ds2E5UHGM~Anwj<1Lk8tDi&uc{g7N%`t zFdh&OKwB6PifWVd5_M$%Rh HU)b7{U}2#W#qf0`Apf5>M__ENb)BES@X z@V%X>)%u=tHChV8e |mdGtZux&^YeKU2V-MOF!A3-CL*hGVX%`JfsQi~~H zlS05brO^|TZ`4h^$Q(ik`0 NIyzlR^=iLJ`R$jm{rwc<5| IoFo$Pd}>k%Q4qo1hWDY@O=Eji3$$=ZS1bJ@Mw#0TV}U 89l&3w z3@JX@SnVizXoIJZ4l?$FsT4C2L5%3ga{-eO4UHfPKnxEl%K3Z+$iYKKWlDXE9Yp62 z`Fd4|&!Ymj8_+&I7`RnwIwxK$(*|QhR>HONtqPB;c{_3YpI2b&ROFYJ(ElgJ%E9Yj zY_kiLLEkR;RoT`uYyePW{0c)`(x}=8lIH%V@+M-sQp_bLkemQsnlH(e3V#pL&$4Qs z!nBa)V?-wV(>rDOTZp@J&;FBg+~GB%`}rRiD7n~*&0pD>R*Uj}kvUx~$pN={SyG_+ z%nTrMQ+d-1O{~O! 1*5i|@sy-+k@6P@KMAIg6vrwljy={Q;<+ zUagBeoZ{=dLu~Io;x!X_{?^W(FLS7RDS-wmUod&fz4>wTi)n3_`s);8@eITeKd1F^ z!p0}jI(NBOy5+EQ4r=W8L^RWe9s>xn_2RWYCXj00UiIHFBO-oE0MG!8Igi@=JT}vP zTw0d9Q4?cin`8e)!Q7?$O?XCnmXY|mzFolX@{QUFT0tV?oDjd%yKe=#SM*GQwi=_2 zbV_^4vKAhq>UX(>;(75;iYPc@ISCK}Sc{F$U1#GTN3e&mx7EIDu}u4O{8Cdntk4~- zRm%8)aDVW!_D1C$l19$A)alc_|2|bZfSmQps;9IpL^wip%6*TUBxEk8%@rrF;5~Gc zKotC%ay8AGBSKj7dB2$6YF 15w` zdzP%j$-HsJAe%!s7HI2FOIA->dBWu1m #(Yo)9i%YZr^qj3(olF==345cg-Eo_i$qW1a-xVwLKl zENeU)8X%|v5f#sGx9h%WyKrTm=-4nKuK|Q8Uh20Sur fv=ZO*;*n3bveK$Q A3{Xp8|x#FXu0jYkmrESLO89o#^O)+KU8F7eNDjSZ) zpVfk>?i{Gm5luO@m-LLU92qb7Jyj4R7|01oS1GHi zV0<=GnYb}A YY-rF|O&gUscR1zVl=f0s`JrA-J|XWv1}+h+i7b1Ldr@T;0 HnYRA zl>*$q0MO{A8VzCq?)fyoeG6dXWitd=7vP{r=0qt?=9T8$T!AB!jix5Lu;v72Db|QK zxr2%J1H>b}uK@pC2f}A+ffG-M#@|6#Gve`%N=AQ`h+Zj@15A0_VFNf6YFvs;MYp`g zE#@whFL6wtLorN`k}h<;3Rb!TAjWcLWE1w&eyGEzIT6e$8SAku&C5?n%Yq9l+Mo*H zK4OUIo(8k48;qccD_PQ$J9?3p|4n1nry!E?RVM;WtFD-Fgq2nQ@JsrIc&nvAn{1|J zrcj9Ga8LpHhmBaVj*-%gxK5EnneIi0*UZaiLqLy6#~ZLQe#BVp=I;5~0e<*$3C-G` zC`Q7(3d^NGf{WV^9N=DsX93#E?DZ-`y|E%XmB8Cv1rSs>$jZ!lC{12jY}+L1mTitK zKb}SAEJWM6zt+-K1e={@?rNq_CfQ(OTwp^Dgp>QqNv@Y`cJINIzeQOP-2g?_VnEKE zDTw2X`h;a*{an~`sem;zSFPrR#B=P7zm*zqE;g}pHDaz5=95&gna3pcrqXP7Tehq> zenm^*ilgtFGL!31nM_sU>;|lyzL)|D;tgZ+_<~<)5^2DFWBC1vhSKncX)Zj-7@k0G zXciQxUd!&WFmJnW*2Gl&sz$nCs^KA0X6C=7V)4}k(Z|dm8!A6P># -2ozK;eR z^HTFTqEj0lQs6uPCDGTvdmhpIGA>QirBVKPxTN9L5HACB%h<6V0^7yd{Zm`Bds;GA z`%~aqmk8r$cp*me7{ z-AYMe=*%kulFvV1ADYM$Y>@xzLSLZ9wi&tbOt0bwJtJELeYsByXG%(7-`tn`MZJYU zN|&)))KrB$=Xhi4mPQsb5olFha{&)B*`+|1Q?|BTKjT|Uq@Ic#zq=RtHeU}Oye#@K zFZ Az&79PG5znQd* zSIe;sO~wduGra5YSnvZPh;bdFD?`aoqEN%Ji=8J=tbzKtUc{o)55q1eXVsKk7!f!~ zF=#n<=u0qTvB+lb ^dp;;%88mK72H28&8tAfRJjDu%(qM&87mDzaKN-zVQ}OT8w32=BY{{-! zgjUDVW)DJU+-NV_XKF&Z-7O8IW~Y|IYrrDioqaiW?)6&;{vpCBHl*nO|N-Lx$cDN+cbRtn&?d4tQV95v@R&P3t3}nH(+qsZS!7PuqV;$VXIo zZ>Fn%9EQy46Cgx9BbHZU{*#mXSICJk-Bj8`wWKrs!)0&B;@^9|xTGda#y63MX-EsR zs0$T_({}xA39*-9t&C#9R}l}$eAX!&pu+`?l-%7ykIVL9YH$yh-OF1^86#Jq#4YcD zQESxv+K(yC0iIqM@4xm}4^;v`4%d9W6QDDnjhzd&jh!!}|2j7tct14czU=;2iQ9ug z0ma(I0UX(vT)7}Fb~`Htg!LkE>+7^J9a`Q!r|}!hV}VXzrb-^u6{L{(tc7wkj3pFw zuC8zo>dm|GCDp?19Iom-0U^V56wVeHK=100VdrHqwsN#kz|glfKaDZ%TiTv3g6Vx# zN9gVKoLefgz~`lN50X3hH#PUN&wXUby(yxvt}2qnM!lQr1Nm<8$t;cjph|vI_<=Ei zpZB#il^l@5XFV{i;55|NF45ooU=_jy+8BR-N_o>TTu48Evw#s%?}vCUn6z@9^=n#J zf`u);ztp|F){&*lu~&j3D_?>J694x0w $Oz2Zwgg%tMX~C>Zcmn+|4IfukhPl_YpoJySE2jh$*n1l-$aa1W|ZZ_ru>^ z7OoCJJ z;q7=_7)RU``t$RDoNj_B!YY0e1?e>hfaLqEFeU&%Uxh-TfD?e9MdKL|XGiJqR16&m z30=PIec-B}EW)Zgqx@ZLDp!WtG~yt*#x$MZ%5SuOqZG&l1Ta26-^BqCSfZ5iI%Y57 zF;XNm{?kusn`D@WVapU1^E!t?T(#LRb1jZR`o{)gmCKC|VsG^Ca^34MCXHei3{rp& zXc2*QVfg5>*_sGt osMc5LjC7{!WpJ*#ypOrFhXi9M`aE`qAF30O z-V11hP>~ &WaC39+@?5z9;NRV8%SbWDW@jl#T+uju{;m$ z6~RzeG34XO1!j2Jbxoz-Y%qhG%Ohh(DW2D`2kAYe8_<0aCg__l+$xB*_9sTIKGNW( zJBtTT4nlzL><}3}Z~UBt#IV7ljYcZxY+BRBYV7qC9b1t4&Z$_D$%j+Of~kQ KzSxI9& zGZf2e@DW>D7-Hq7m`QB`E3b?tBrOpok-6j1n+_H5SthA-O#I|{e82H$W&vM#jf**e zyhj9G7sL>4^PI!E)SrHL2cWHECxBhME?O5Z8lWc$=i@y9F8z(?Y^=AslUSI-clJ{Y zl0~_QdSbP5$3av5f28kOnp`76+1sv{M S<>9hh(b zu|g^0h#Uhg{I{qBi{i}+>