「Pythonによるスクレイピング＆機械学習」を読んで

こちらの本の補足情報を書いています。

Pythonによるスクレイピング&機械学習 開発テクニック BeautifulSoup,scikit-learn,TensorFlowを使ってみよう

公式の正誤情報

４章あたり

.sh

pip3 install sklearn scipy pandas
python3 xor-train.py

deprecation

.diff

diff iris-train2.py{.bk,}
2c2
< from sklearn import svm, metrics, cross_validation
---
> from sklearn import svm, metrics, model_selection
13c13
<     cross_validation.train_test_split(csv_data, csv_label)
---
>     model_selection.train_test_split(csv_data, csv_label)

.diff

diff mnist-train.py{bk,}
1c1
< from sklearn import cross_validation, svm, metrics
---
> from sklearn import svm, metrics

言語判定

pip3 install matplotlib numpy pandas
export MPLBACKEND="agg"
python3 lang-plot.py
open lang-plot.png

BMI判定

.diff

diff bmi-test.py{.bk,}
1c1
< from sklearn import cross_validation, svm, metrics
---
> from sklearn import model_selection, svm, metrics
16c16
<     cross_validation.train_test_split(wh, label)
---
>     model_selection.train_test_split(wh, label)

5章

.diff

diff mnist-deep.py{.bk,}                                                                                                  
91c91
<     sess.run(tf.initialize_all_variables())
---
>     sess.run(tf.global_variables_initializer())
93c93
<     tw = tf.train.SummaryWriter('log_dir', graph=sess.graph)
---
>     tw = tf.summary.FileWriter('log_dir', graph=sess.graph)

7章

画像OCR

.diff

diff font_draw.py{.bk,} 
26c26
< if not os.path.exists("./image/num"): os.makedires("./image/num")
---
> if not os.path.exists("./image/num"): os.makedirs("./image/num")

$ diff caltech101_makedata.py{.bk,}
1c1
< from sklearn import cross_validation
---
> from sklearn import model_selection, svm, metrics
40c40
<     cross_validation.train_test_split(X, Y)
---
>     model_selection.train_test_split(X, Y)

もしOpenCV4を使用しているならエラーが出てしまうので修正

$ python ocr1.py
Traceback (most recent call last):
  File "ocr1.py", line 18, in <module>
    x, y, w, h = cv2.boundingRect(cnt) # --- (※5)
cv2.error: OpenCV(4.1.0) /Users/travis/build/skvark/opencv-python/opencv/modules/imgproc/src/shapedescr.cpp:743: error: (-215:Assertion failed) npoints >= 0 && (depth == CV_32F || depth == CV_32S) in function 'pointSetBoundingRect'

$ diff ocr1.py{.bk,}
14c14
<     thresh, cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)[1]
---
>     thresh, cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)[0]

$ diff ocr3.py{.bk,}
19c19
<     cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)[1]
---
>     cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)[0]

Caltech101データセット