packtpublishing/deep-learning-with-keras

Code repository for Deep Learning with Keras published by Packt

deep-learning-with-keras's Issues

Semi-supervised learning GANs for classification

Hello,

I have been reading the book to try to understand the GANS. I would like to try it (GANs) for classification.

I have been looking to this code :https://github.com/PacktPublishing/Deep-Learning-with-Keras/blob/master/Chapter04/example_gan_convolutional.py

To try to adapt it to a GAN with N+1 outputs being N the number of classes and +1 for the forge output.
Some things I dont understand on the example, like the output generation gan_targets(xtest.shape[0]). ... How exactly can the GAN architecture test if the output is fake or not ?

And to adapt this example, to make a Semi-supervised learning for classification, how should we incorporate the output layer of 10 with soft max and construct the outputs align with the inputs (and generated -fake- inputs with label ) ?

I would be grateful if there was some example, based on this one, were we could see GANs performing classification ?

Thanks in advance,
Rui

Dropout layer error in transfer_glove_embeddings.py

Hi all!

I am trying to replicate the model of transfer_glove_embeddings.py. When I try to execute the model I gel the following error:

`---------------------------------------------------------------------------
TypeError Traceback (most recent call last)
in ()
8 weights=[embedding_weights],
9 trainable=True))
---> 10 model.add(SpatialDropout1D(Dropout(0.2)))
11 model.add(Conv1D(filters=NUM_FILTERS, kernel_size=NUM_WORDS,
12 activation="relu"))

~/.virtualenvs/negative-scoring/lib/python3.6/site-packages/keras/legacy/interfaces.py in wrapper(*args, **kwargs)
89 warnings.warn('Update your ' + object_name + 90 ' call to the Keras 2 API: ' + signature, stacklevel=2)
---> 91 return func(*args, **kwargs)
92 wrapper._original_function = func
93 return wrapper

~/.virtualenvs/negative-scoring/lib/python3.6/site-packages/keras/layers/core.py in init(self, rate, **kwargs)
162 @interfaces.legacy_spatialdropout1d_support
163 def init(self, rate, **kwargs):
--> 164 super(SpatialDropout1D, self).init(rate, **kwargs)
165 self.input_spec = InputSpec(ndim=3)
166

~/.virtualenvs/negative-scoring/lib/python3.6/site-packages/keras/legacy/interfaces.py in wrapper(*args, **kwargs)
89 warnings.warn('Update your ' + object_name + 90 ' call to the Keras 2 API: ' + signature, stacklevel=2)
---> 91 return func(*args, **kwargs)
92 wrapper._original_function = func
93 return wrapper

~/.virtualenvs/negative-scoring/lib/python3.6/site-packages/keras/layers/core.py in init(self, rate, noise_shape, seed, **kwargs)
99 def init(self, rate, noise_shape=None, seed=None, **kwargs):
100 super(Dropout, self).init(**kwargs)
--> 101 self.rate = min(1., max(0., rate))
102 self.noise_shape = noise_shape
103 self.seed = seed

TypeError: '>' not supported between instances of 'Dropout' and 'float'`

Any ideas? thanks in advanced

Serialise and reuse the model using a random sentence

Hi!
First of all, I would like to say that I'm enjoying the book a lot so far! Good job!

What I would like to know is:

1. how to serialise/save the model (SOLVED)
   (found it! at https://keras.io/getting-started/faq/#how-can-i-use-pre-trained-models-in-keras)

2.how to load and re-use the model passing a random sentence.
Thank you in advance!

TypeError: a bytes-like object is required, not 'str'

ipython-input-11-cde6afe50c78 in module()
5 ftrain=open(os.path.join(Data_dir,"training.txt"),'rb')
6 for line in ftrain:
----> 7 label, sentence = line.strip().split("\t")
8 words=nltk.word_tokenize(sentence.decode().lower())
9 if len(words)>maxlen:

I am using Python3.6

2.2.5 Key Attributes Page 33

It seems there is an error on line which start with

So what we have here is a 3D tensor of 8-bit integers. More precisely, it's an array of 60,000 matrices of 28 x 8 integers.

This should be 28 X 28 integrers.

https://github.com/PacktPublishing/Deep-Learning-with-Keras/blob/master/Chapter07/sent-thoughts-rnn.py

Thank you for your book, I made the first chapter in Chinese notes, the level is not high, welcome comments.

Thank you for your book, I made the first chapter in Chinese notes, the level is not high, welcome comments.

leNet not work with keras 2.3.1

Chapter03/keras_LeNet.py
line 58: AttributeError: module 'keras.backend' has no attribute 'set_image_dim_ordering'
line 28: ValueError: Negative dimension size caused by subtracting 2 from 1 for 'max_pooling2d_1/MaxPool' (op: 'MaxPool') with input shapes: [?,
1,28,20].

code bug in keras_CIFAR10_simple.py

Hi, authors,

In this line, it should be input_shape=(IMG_CHANNELS, IMG_ROWS, IMG_COLS) as the CIFAR 10 is loaded in the shape (num, channels, rows, columns)

nothing

nothing

could you please provide dataset umich-sentiment-train.txt as well?

Hello, I download the dataset from Kaggle, but it does not work. seem your dataset is . binary?

DP-Keras

where is the link of data?

how to download the book'data?