-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathall.html
executable file
·897 lines (614 loc) · 29.8 KB
/
all.html
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
<!doctype html>
<html class="no-js" lang="en">
<head>
<meta charset="utf-8" />
<meta name="viewport" content="width=device-width, initial-scale=1.0" />
<title>
雪地
</title>
<link href="atom.xml" rel="alternate" title="雪地" type="application/atom+xml">
<link rel="stylesheet" href="asset/css/foundation.min.css" />
<link rel="stylesheet" href="asset/css/docs.css" />
<link rel="icon" href="asset/img/favicon.ico" />
<script src="asset/js/vendor/modernizr.js"></script>
<script src="asset/js/vendor/jquery.js"></script>
<script src="asset/highlightjs/highlight.pack.js"></script>
<link href="asset/highlightjs/styles/github.css" media="screen, projection" rel="stylesheet" type="text/css">
<script>hljs.initHighlightingOnLoad();</script>
<script type="text/javascript">
function before_search(){
var searchVal = 'site:yinzo.github.io ' + document.getElementById('search_input').value;
document.getElementById('search_q').value = searchVal;
return true;
}
</script>
</head>
<body class="antialiased hide-extras">
<div class="marketing off-canvas-wrap" data-offcanvas>
<div class="inner-wrap">
<nav class="top-bar docs-bar hide-for-small" data-topbar>
<section class="top-bar-section">
<div class="row">
<div style="position: relative;width:100%;"><div style="position: absolute; width:100%;">
<ul id="main-menu" class="left">
<li id="menu_item_index"><a href="index.html">Blog</a></li>
<li id="menu_item_archives"><a href="archives.html">Archives</a></li>
<li id="menu_item_about"><a href="http://yinz.xyz/">Home</a></li>
</ul>
<ul class="right" id="search-wrap">
<li>
<form target="_blank" onsubmit="return before_search();" action="http://google.com/search" method="get">
<input type="hidden" id="search_q" name="q" value="" />
<input tabindex="1" type="search" id="search_input" placeholder="Search"/>
</form>
</li>
</ul>
</div></div>
</div>
</section>
</nav>
<nav class="tab-bar show-for-small">
<a href="javascript:void(0)" class="left-off-canvas-toggle menu-icon">
<span> 雪地</span>
</a>
</nav>
<aside class="left-off-canvas-menu">
<ul class="off-canvas-list">
<li><a href="index.html">Blog</a></li>
<li><a href="archives.html">Archives</a></li>
<li><a href="http://yinz.xyz/">Home</a></li>
<li><label>Categories</label></li>
<li><a href="Security%20Info.html">Security Info</a></li>
<li><a href="Adversary%20Learning.html">Adversary Learning</a></li>
<li><a href="TCPIP.html">TCP/IP</a></li>
<li><a href="Pattern%20Recognition.html">Pattern Recognition</a></li>
<li><a href="Python.html">Python</a></li>
<li><a href="OS.html">OS</a></li>
<li><a href="Deep%20Learning.html">Deep Learning</a></li>
<li><a href="Machine%20Learning.html">Machine Learning</a></li>
</ul>
</aside>
<a class="exit-off-canvas" href="#"></a>
<section id="main-content" role="main" class="scroll-container">
<script type="text/javascript">
$(function(){
$('#menu_item_index').addClass('is_active');
});
</script>
<div class="row">
<div class="large-8 medium-8 columns">
<div class="markdown-body home-categories">
<div class="article">
<a class="clearlink" href="14968173531750.html">
<h1>CS229 学习笔记 Part3</h1>
<div class="a-content">
<div class="a-content-text">
<ul>
<li>
<a href="#toc_0">SVM</a>
<ul>
<li>
<a href="#toc_1">函数间隔和几何间隔 (Functional and geometric margin)</a>
</li>
<li>
<a href="#toc_2">最优间隔分类器 (Optimal margin classifier)</a>
</li>
<li>
<a href="#toc_3">拉格朗日对偶问题</a>
<ul>
<li>
<a href="#toc_4">原始问题</a>
</li>
<li>
<a href="#toc_5">对偶问题</a>
</li>
<li>
<a href="#toc_6">KKT 条件</a>
</li>
</ul>
</li>
<li>
<a href="#toc_7">回到最优间隔分类器问题</a>
</li>
<li>
<a href="#toc_8">核 (Kernels)</a>
</li>
<li>
<a href="#toc_9">正则化</a>
</li>
<li>
<a href="#toc_10">SMO 算法</a>
<ul>
<li>
<a href="#toc_11">坐标上升算法 (Coordinate ascent)</a>
</li>
<li>
<a href="#toc_12">SMO</a>
</li>
</ul>
</li>
</ul>
</li>
</ul>
<h2 id="toc_0">SVM</h2>
<p>CS229 对于 SVM 的理论解释是我学习到的最详细也是最好的一份资料了，对比对象有周志华《机器学习》、《机器学习实战》、Coursera 上的 Machine Learning 等。相当推荐学习 CS229。</p>
<p>分类间隔 (Margin) 和 SVM 的优化目标『最大化分类间隔』这里就不多说了，很好理解，主要还是记录 CS229 中学到的新内容。一个数据点离分类边界 (decision boundary) 越远，则确信度越高。我们的优化目标也相当于寻找一个远离所有数据点的分类边界，当然，前提是这个分类边界得到的分类都正确。</p>
<p>SVM 的一些特殊定义也提及一下，</p>
<ul>
<li>\(y\) 的取值不是 \(\{0,1\}\) 而是 \(\{-1,1\}\)。</li>
<li>假设函数 \(h_{w,b}(x) = g(w^Tx+b)\) 中，我们把截距项单独写出来，便与后续的计算。</li>
<li>我们的分类器输出结果会直接是 1 或 -1，不像 Logistic 回归那样先输出 \(y\) 是某一类的概率。</li>
</ul>
</div>
</div>
</a>
<div class="read-more clearfix">
<div class="more-left left">
<a href="14968173531750.html">Read more</a>
<span class="date">2017/6/7 14:35 下午</span>
<span>posted in </span>
<span class="posted-in"><a href='Machine%20Learning.html'>Machine Learning</a></span>
</div>
<div class="more-right right">
<span class="comments">
</span>
</div>
</div>
</div><!-- article -->
<div class="article">
<a class="clearlink" href="14965964854250.html">
<h1>CS229 学习笔记 Part2</h1>
<div class="a-content">
<div class="a-content-text">
<ul>
<li>
<a href="#toc_0">判别式和生成式</a>
<ul>
<li>
<a href="#toc_1">判别式</a>
</li>
<li>
<a href="#toc_2">生成式</a>
</li>
</ul>
</li>
<li>
<a href="#toc_3">Gaussian discriminant analysis</a>
</li>
<li>
<a href="#toc_4">讨论：GDA 和 Logistic 回归</a>
</li>
<li>
<a href="#toc_5">朴素贝叶斯</a>
<ul>
<li>
<a href="#toc_6">拉普拉斯平滑</a>
</li>
<li>
<a href="#toc_7">用于文本分类的事件模型</a>
</li>
</ul>
</li>
</ul>
<h2 id="toc_0">判别式和生成式</h2>
<p>对于一个分类任务，判别式和生成式分别代表了两种不同的思路：</p>
<h3 id="toc_1">判别式</h3>
<p>通过直接从输入数据中学习，得到一个『特定输入对应的实际类别』的概率模型，模型的参数为 \(\theta\) 。即学习建模 \(p(y\mid x)\)</p>
<h3 id="toc_2">生成式</h3>
<p>通过对每一个类进行建模，然后就可以通过条件概率算出输入的数据更可能由哪一类生成。即学习建模 \(p(x\mid y)\) 和 \(p(y)\) ，然后计算 \[\arg\max\limits_y\frac{p(x \mid y)p(y)}{p(x)}\]</p>
<p>并且实际计算中，分母 \(p(x)\) 并不会影响各个类别概率的排序，所以最终简化成 \[\arg\max\limits_y p(x \mid y)p(y)\]</p>
</div>
</div>
</a>
<div class="read-more clearfix">
<div class="more-left left">
<a href="14965964854250.html">Read more</a>
<span class="date">2017/6/5 1:14 上午</span>
<span>posted in </span>
<span class="posted-in"><a href='Machine%20Learning.html'>Machine Learning</a></span>
</div>
<div class="more-right right">
<span class="comments">
</span>
</div>
</div>
</div><!-- article -->
<div class="article">
<a class="clearlink" href="14946020792948.html">
<h1>CS229 学习笔记 Part 1</h1>
<div class="a-content">
<div class="a-content-text">
<p>此笔记为我的 CS229 的学习笔记之一，由 Andrew Ng 的 CS229 Lecture notes 和 课堂录像整理而来。用于记录所学到的内容。记录顺序重新编排过，并非是课程原本的教学顺序，并且省略了课程中的一些推导过程，所以适合学习后整理备忘使用，不适合用于同步辅助学习。</p>
<ul>
<li>
<a href="#toc_0">广义线性模型 GLM (Generalized Linear Models)</a>
<ul>
<li>
<a href="#toc_1">具体步骤</a>
</li>
</ul>
</li>
<li>
<a href="#toc_2">优化方法</a>
<ul>
<li>
<a href="#toc_3">梯度下降法</a>
</li>
<li>
<a href="#toc_4">牛顿法</a>
</li>
</ul>
</li>
<li>
<a href="#toc_5">Linear Regression</a>
<ul>
<li>
<a href="#toc_6">Locally weighted linear regression</a>
</li>
</ul>
</li>
</ul>
<h2 id="toc_0">广义线性模型 GLM (Generalized Linear Models)</h2>
<p>广义线性模型是所学到的 Linear Regression 以及 Logistic Regression 的推广形式（更准确的说，这两种模型都属于 GLM 的特殊情况）。它有三个关键假设(Assumptions)构成:</p>
<ol>
<li>\(y \mid x;\theta\sim ExponentialFamily(\eta)\) ：对于固定的参数 \(\theta\) 以及给定 \(x\)， \(y\) 的分布服从某一指数分布族（如高斯分布、伯努利分布、Softmax分布）</li>
<li>对于给定的 \(x\) ，目标是预测 \(T(y)\) 的值。换一种说法就是，我们定义假设函数 \(h(x) = E[y\mid x]\)</li>
<li>natural parameter \(\eta\) 和 输入 \(x\) 是线性相关的， \(\eta = \theta^ \mathrm{ T } x\) （其中，当输入 \(x\) 和 \(\eta\) 是向量的时候， \(\eta_i = \theta_i^ \mathrm{T}x\)）</li>
</ol>
</div>
</div>
</a>
<div class="read-more clearfix">
<div class="more-left left">
<a href="14946020792948.html">Read more</a>
<span class="date">2017/5/12 23:14 下午</span>
<span>posted in </span>
<span class="posted-in"><a href='Machine%20Learning.html'>Machine Learning</a></span>
</div>
<div class="more-right right">
<span class="comments">
</span>
</div>
</div>
</div><!-- article -->
<div class="article">
<a class="clearlink" href="14883590547961.html">
<h1>原始模型优化笔记</h1>
<div class="a-content">
<div class="a-content-text">
<p>对于原始弹幕分类CNN模型进行优化。</p>
<h2 id="toc_0">修改 word2vec model 的 vector size</h2>
<ul>
<li>400:
Nice at epoch 38, validation acc 96.56%</li>
<li>200:
Nice at epoch 37, validation acc 95.22%</li>
<li>100:
Nice at epoch 34, validation acc 94.78%
单轮训练时间与50维相近，测试样例测试耗时 0.92secs</li>
<li>50:
Nice at epoch 40, validation acc 94.39%
单轮训练时间在7秒左右，测试样例(av 8365806)测试耗时 0.7secs</li>
</ul>
<h2 id="toc_1">尝试加入dropout</h2>
<p>在两个 conv 层之间和两个 fc 层之间各加入了一个 \(p=0.5\) 的 dropout</p>
<p>40 epoch 时只有 89.1 acc， 和预想的一样，会导致 达到最佳效果的 epoch 数上升。</p>
<p>用了 dropout 后一个很明显的变化是，原本训练过程中通常是train acc 高于 validation acc，现在通常是 validation acc 高于 train acc，训练后期才基本持平或反超</p>
<p>vector在 epoch 90 左右 达到了96.50%上下的 acc，最终在epoch 300 以上能达到 97.10% 左右的 acc</p>
<p><img src="media/14879250453025/%E5%B1%8F%E5%B9%95%E5%BF%AB%E7%85%A7%202017-02-24%2022.15.27.png" alt="屏幕快照 2017-02-24 22.15.27"/></p>
</div>
</div>
</a>
<div class="read-more clearfix">
<div class="more-left left">
<a href="14883590547961.html">Read more</a>
<span class="date">2017/3/1 17:4 下午</span>
<span>posted in </span>
<span class="posted-in"><a href='Deep%20Learning.html'>Deep Learning</a></span>
</div>
<div class="more-right right">
<span class="comments">
</span>
</div>
</div>
</div><!-- article -->
<div class="article">
<a class="clearlink" href="14863637393852.html">
<h1>低素质弹幕分类器的CNN实现</h1>
<div class="a-content">
<div class="a-content-text">
<h2 id="toc_0">整体架构</h2>
<p>对于一条弹幕，首先进行分词，然后通过 word2vec 转换为词向量，再填充至固定长度，作为卷积神经网络的输入。</p>
<p>卷积神经网络的结构如下：</p>
<pre><code class="language-python">model = Sequential()
model.add(Convolution1D(100, 4, border_mode='valid', input_shape=(100, word_model.vector_size)))
model.add(Activation('relu'))
model.add(Convolution1D(100, 4, border_mode='valid', input_shape=(100, word_model.vector_size)))
model.add(Activation('relu'))
model.add(Flatten())
model.add(Dense(100, activation='relu'))
model.add(Dense(2, activation='softmax'))
model.compile(loss='categorical_crossentropy',
optimizer='adam',
metrics=['accuracy']
)
</code></pre>
<p>最终输出为2位的 categorical result，直接使用第一项，即骂人弹幕的概率作为输出。</p>
<p>然后通过代理，在弹幕服务器与播放器之间插入一层，实现弹幕的分类与屏蔽。最终实现了有效的骂人弹幕自动屏蔽，但是误伤的情况依然存在。</p>
<h2 id="toc_1">搭建过程</h2>
</div>
</div>
</a>
<div class="read-more clearfix">
<div class="more-left left">
<a href="14863637393852.html">Read more</a>
<span class="date">2017/2/6 14:48 下午</span>
<span>posted in </span>
<span class="posted-in"><a href='Deep%20Learning.html'>Deep Learning</a></span>
</div>
<div class="more-right right">
<span class="comments">
</span>
</div>
</div>
</div><!-- article -->
<div class="article">
<a class="clearlink" href="14863637259966.html">
<h1>低素质弹幕分类器 CNN 训练笔记</h1>
<div class="a-content">
<div class="a-content-text">
<p>一开始使用这个结构，迭代10次。</p>
<pre><code class="language-python">model = Sequential()
model.add(Convolution1D(100, 4, border_mode='valid', input_shape=(100, word_model.vector_size)))
model.add(Activation('relu'))
model.add(Convolution1D(5, 4, border_mode='valid'))
model.add(Activation('relu'))
model.add(Flatten())
model.add(Dense(32, activation='relu'))
model.add(Dense(2, activation='softmax'))
model.compile(loss='categorical_crossentropy',
optimizer='adam',
metrics=['accuracy']
)
</code></pre>
<p>完成训练后，乍一看准确率很高，结果 print 出来看一下，低素质弹幕完全没有被过滤，完全是将分类全部丢给 positive 达到的高准确率 (0.98) 的确是 meaningless classification<br/>
并且这个结果在loss里看得很清楚，loss一直是处于15+的</p>
</div>
</div>
</a>
<div class="read-more clearfix">
<div class="more-left left">
<a href="14863637259966.html">Read more</a>
<span class="date">2017/2/6 14:48 下午</span>
<span>posted in </span>
<span class="posted-in"><a href='Deep%20Learning.html'>Deep Learning</a></span>
</div>
<div class="more-right right">
<span class="comments">
</span>
</div>
</div>
</div><!-- article -->
<div class="article">
<a class="clearlink" href="14828521678770.html">
<h1>在 Linode 上编译 hybla 模块</h1>
<div class="a-content">
<div class="a-content-text">
<p>考试复习期间，不想复习，搞搞其他东西散散心，发现shadowsocks 有关于 TCP Fast Open 的更新，看说明的过程还发现了官方的速度优化指南，尝试优化自己的 ss 速度，然而报错。</p>
<p>执行到这一步时运行出错</p>
<pre><code class="language-sh">sysctl --system
</code></pre>
<pre><code class="language-sh">...
sysctl: setting key "net.ipv4.tcp_congestion_control": No such file or directory
net.ipv4.tcp_congestion_control = hybla
* Applying /etc/sysctl.conf ...
</code></pre>
</div>
</div>
</a>
<div class="read-more clearfix">
<div class="more-left left">
<a href="14828521678770.html">Read more</a>
<span class="date">2016/12/27 23:22 下午</span>
<span>posted in </span>
<span class="posted-in"><a href='OS.html'>OS</a></span>
</div>
<div class="more-right right">
<span class="comments">
</span>
</div>
</div>
</div><!-- article -->
<div class="article">
<a class="clearlink" href="14696108056583.html">
<h1>无需手动缓存长度</h1>
<div class="a-content">
<div class="a-content-text">
<p>今天在阅读《机器学习实战》的时候，看到了这样一句描述</p>
<p><img src="media/14696108056583/IMG_0629.png" alt="IMG_0629"/></p>
<p>其中数据集是一个List。看到这里说到为了提高代码效率，特地开了一个变量来保存其长度。</p>
</div>
</div>
</a>
<div class="read-more clearfix">
<div class="more-left left">
<a href="14696108056583.html">Read more</a>
<span class="date">2016/7/27 17:13 下午</span>
<span>posted in </span>
<span class="posted-in"><a href='Python.html'>Python</a></span>
</div>
<div class="more-right right">
<span class="comments">
</span>
</div>
</div>
</div><!-- article -->
<div class="article">
<a class="clearlink" href="14685108901937.html">
<h1>使用 Python 装饰器来将普通函数加入任务队列</h1>
<div class="a-content">
<div class="a-content-text">
<p>对于操作『函数对象』来说，使用 Python 装饰器是一种非常优雅，非常 Pythonic 的一个方式。而在这篇文章中，对于任何一个普通的函数，只需要在函数定义前加一个装饰器调用，即可使得这一函数被调用时自动加入特定的任务队列，成为异步调用，而不会阻塞主线程。</p>
</div>
</div>
</a>
<div class="read-more clearfix">
<div class="more-left left">
<a href="14685108901937.html">Read more</a>
<span class="date">2016/7/14 23:41 下午</span>
<span>posted in </span>
<span class="posted-in"><a href='Python.html'>Python</a></span>
</div>
<div class="more-right right">
<span class="comments">
</span>
</div>
</div>
</div><!-- article -->
<div class="article">
<a class="clearlink" href="14610807170718.html">
<h1>关于Python的logging模块初始化无效的一个小坑</h1>
<div class="a-content">
<div class="a-content-text">
<p>众所周知，logging模块是一个非常方便好用的日志输出模块。但是最近的使用发现了一个小坑，记录一下，避免再次踩坑。</p>
<p>一般使用logging模块都会对其进行初始化，使用以下代码：</p>
<pre><code> log_format = '[%(levelname)s] %(asctime)s %(filename)s line %(lineno)d: %(message)s'
date_fmt = '%a, %d %b %Y %H:%M:%S'
logging.basicConfig(
format=log_format,
datefmt=date_fmt,
level=log_level,
)
</code></pre>
</div>
</div>
</a>
<div class="read-more clearfix">
<div class="more-left left">
<a href="14610807170718.html">Read more</a>
<span class="date">2016/4/19 23:45 下午</span>
<span>posted in </span>
<span class="posted-in"><a href='Python.html'>Python</a></span>
</div>
<div class="more-right right">
<span class="comments">
</span>
</div>
</div>
</div><!-- article -->
<div class="row">
<div class="large-6 columns">
<p class="text-left" style="padding-top:25px;">
</p>
</div>
<div class="large-6 columns">
<p class="text-right" style="padding-top:25px;">
<a href="all_1.html">» Next Page</a>
</p>
</div>
</div>
</div>
</div><!-- large 8 -->
<div class="large-4 medium-4 columns">
<div class="hide-for-small">
<div id="sidebar" class="sidebar">
<div id="site-info" class="site-info">
<div class="site-a-logo"><img src="asset/img/3.png" /></div>
<h1>雪地</h1>
<div class="site-des"></div>
<div class="social">
<a class="github" target="_blank" href="https://github.com/Yinzo" title="GitHub">GitHub</a>
<a class="email" href="mailto:[email protected]" title="Email">Email</a>
<a class="rss" href="atom.xml" title="RSS">RSS</a>
</div>
</div>
<div id="site-categories" class="side-item ">
<div class="side-header">
<h2>Categories</h2>
</div>
<div class="side-content">
<p class="cat-list">
<a href="Security%20Info.html"><strong>Security Info</strong></a>
<a href="Adversary%20Learning.html"><strong>Adversary Learning</strong></a>
<a href="TCPIP.html"><strong>TCP/IP</strong></a>
<a href="Pattern%20Recognition.html"><strong>Pattern Recognition</strong></a>
<a href="Python.html"><strong>Python</strong></a>
<a href="OS.html"><strong>OS</strong></a>
<a href="Deep%20Learning.html"><strong>Deep Learning</strong></a>
<a href="Machine%20Learning.html"><strong>Machine Learning</strong></a>
</p>
</div>
</div>
<div id="site-categories" class="side-item">
<div class="side-header">
<h2>Recent Posts</h2>
</div>
<div class="side-content">
<ul class="posts-list">
<li class="post">
<a href="14968173531750.html">CS229 学习笔记 Part3</a>
</li>
<li class="post">
<a href="14965964854250.html">CS229 学习笔记 Part2</a>
</li>
<li class="post">
<a href="14946020792948.html">CS229 学习笔记 Part 1</a>
</li>
<li class="post">
<a href="14883590547961.html">原始模型优化笔记</a>
</li>
<li class="post">
<a href="14863637393852.html">低素质弹幕分类器的CNN实现</a>
</li>
</ul>
</div>
</div>
<div id="site-link" class="side-item">
<div class="side-header">
<h2>Link</h2>
</div>
<div class="side-content">
<p class="link-list">
<a href="http://blog.winkidney.com/">阿毛</a>
</p>
</div>
</div>
</div><!-- sidebar -->
</div><!-- hide for small -->
</div><!-- large 4 -->
</div><!-- row -->
<div class="page-bottom clearfix">
<div class="row">
<p class="copyright">Copyright © 2016
Powered by <a target="_blank" href="http://www.mweb.im">MWeb</a>,
Theme used <a target="_blank" href="http://github.com">GitHub CSS</a>.
Modified by <a target="_blank" href="http://yinz.xyz">Yinzo</a>.</p>
</div>
</div>
</section>
</div>
</div>
<script src="asset/js/foundation.min.js"></script>
<script>
$(document).foundation();
function fixSidebarHeight(){
var w1 = $('.markdown-body').height();
var w2 = $('#sidebar').height();
if (w1 > w2) { $('#sidebar').height(w1); };
}
$(function(){
fixSidebarHeight();
})
$(window).load(function(){
fixSidebarHeight();
});
</script>
<script src="asset/chart/all-min.js"></script><script type="text/javascript">$(function(){ var mwebii=0; var mwebChartEleId = 'mweb-chart-ele-'; $('pre>code').each(function(){ mwebii++; var eleiid = mwebChartEleId+mwebii; if($(this).hasClass('language-sequence')){ var ele = $(this).addClass('nohighlight').parent(); $('<div id="'+eleiid+'"></div>').insertAfter(ele); ele.hide(); var diagram = Diagram.parse($(this).text()); diagram.drawSVG(eleiid,{theme: 'simple'}); }else if($(this).hasClass('language-flow')){ var ele = $(this).addClass('nohighlight').parent(); $('<div id="'+eleiid+'"></div>').insertAfter(ele); ele.hide(); var diagram = flowchart.parse($(this).text()); diagram.drawSVG(eleiid); } });});</script>
<script type="text/javascript" src="https://cdn.mathjax.org/mathjax/latest/MathJax.js?config=TeX-AMS-MML_HTMLorMML"></script><script type="text/x-mathjax-config">MathJax.Hub.Config({TeX: { equationNumbers: { autoNumber: "AMS" } }});</script>
<script src="asset/js/instantclick.min.js" data-no-instant></script>
<script data-no-instant>InstantClick.on('change', function() {
MathJax.Hub.Queue(["Typeset",MathJax.Hub]);
});</script>
<script data-no-instant>InstantClick.init();</script>
</body>
</html>