Errata: July 21, 2020
Unless otherwise noted, all corrections have been made to the pbook and are pending in mobile and livebook versions.
Thank you for purchasing Deep Learning with Python. Please post any errors, other than those listed below, in the book's Author Online Forum.

Copyright Page:

The Development Editor for this book is Christina Taylor, not Toni Arritola, as listed.

Page 6:

In the last line of section 1.1.3, RG should be RGB.

Page 31:

Replace the first three lines of the snippet at the beginning of section 2.2.2 with the following:

>>> x = np.array([12, 3, 6, 14, 7])
>>> x
array([12,3,6,14,7])

Page 33:

In the next to the last line of the first paragraph, the simple expression 28 x 8 integers should be 28 x 28.

Page 42:

In the second line following figure 2.5, replace and with of.

Because the rows of x and the columns of y...

Page 51:

There's a type in the Nesterov momentum example; the + sigm should be a minus sign:

velocity = past_velocity * momentum + learning_rate * gradient

Page 65:

This correction hasn't been made in any format.

In the third line of the second paragraph, the upper case W should be lower case:
...backends support Windows), We don't recommend it.
should be
backends support Windows), we don't recommend it.

Page 74, listing 3.9:

Change the fifth line of code to the following:

epochs = range(1, len(loss_values) + 1)

Page 75, listing 3.10:

Replace the second and third lines of code with the following:

acc = history_dict['acc']
val_acc = history_dict['val_acc']

In line 8, replace Loss with Accuracy:

plt.ylabel('Accuracy')

Page 75, figure 3.8:

The Loss legend should read Accuracy.

Page 81, listing 3.20:

Change the eighth line of code to the following:

plt.ylabel('Accuracy')

Page 82, figure 3.10:

The Loss legend should read Accuracy.

Page 84:

In the last line of section 3.5.7, replace eight with four.

The network is able to cram most of the necessaray information into these four dimensional...

Page 87, figure 3.11:

The label for Folds 2 and 3 in the first partition should read Training, not Validation.

Page 99, figure 4.2:

The label for Folds 2 and 3 in the first partition should read Training, not Validation.

Page 100:

The last line of the first paragraph is missing the word be.

...which can be very expensive.

Page 134:

In the note at the top of the page, the simple expression 148 x 148 to 7 x 7 should be 150 x 150 to 7 x 7.

Page 143:

In the third line of the second paragraph, the word it is missing.

... you can expect it to perform...

Page 161, listing 5.25

Line 8, <1>Its shape is (1, 150, 150, 3), is an annotation that didn't render correctly; that annotation goes with the line immediately following.

Page 173:

In line four of the paragraph following listing 5.40, the simple expression 224 x 244 should be 224 x 224.

Page 178:

In line two of the first paragraph, word should be plural.

...sequences of words or sequences...

Page 187, listing 6.7:

Line 2 is missing the term Embedding:

from keras.layers import Flatten, Dense, Embedding

Page 203:

Line 7 is missing the word and:

...via an activation function and a multiplication operation...

Page 207:

In the first line of paragraph two in section 6.3.1, the word as is missing.

...(such as air temperature, ...

Page 209, listing 6.30:

In line 2, <1> temperature (in degrees Celsius) is an annotation that didn't render correctly.

Page 212, listing 6.34:

Add // batch_size to the last two lines:

val_stseps = (300000 - 200001 - lookback) // batch_size
tes_steps = (len(float_data) - 300001 - lookback) // batch_size

Page 227:

In line four of the first paragraph following listing 6.46, the word eight should be four.

...for the right number of epochs (four) and run it...

Page 249:

In line two of the second paragraph in section 7.2.1, the word in should be is.

...that the validation loss is no longer...

Page 261:

In line two of the Batch renormalization sidebar, the word clears should be clear.

It offers clear benefits...

Page 279:

In the second line of the first bullet point, the word tokens() should be token(s).

...to predict the next token(s)...

Page 296:

In line four of section 8.4, the word implementations shold not be plural.

...alongside implementation details relative to...

Page 297:

In line five of the first paragraph in section 8.4.2, the word vector is singular.

...there may be a smile vector, such that...

In the fourth line from the bottom, the word as should be a.

...turning a male face into a female face, ..

Page 299:

In line five of paragraph two, the word accounting should be account.

...process should be taken into account during encoding...

Page 312:

In the second line of the first bullet point, the word differentiate is misspelled.

In line four of the first bullet point, the word nevers should be never.

...the generator never sees images...

In the caption for figure 8.18, the word row should be column.

...in each column, two images...