Added Random Normal Init and fixed optimizer hyperparameters

[arshagarwal]

In response to the discussions in this pull request, this pull request adds Random normal initializer with mean 0 and std-dev 1, and fixes the optimizer hyper-parameters as per DC-GAN paper.

I am sure there is a more graceful way to do this, but I found this way much easier than to fix code in the same pull request.
Sorry about that!

[github-actions]

Preview

Preview and run these notebook edits with Google Colab:

• site/en/guide/advanced_autodiff.ipynb
• site/en/guide/autodiff.ipynb
• site/en/guide/basic_training_loops.ipynb
• site/en/guide/checkpoint.ipynb
• site/en/guide/data.ipynb
• site/en/guide/data_performance.ipynb
• site/en/guide/distributed_training.ipynb
• site/en/guide/eager.ipynb
• site/en/guide/estimator.ipynb
• site/en/guide/function.ipynb
• site/en/guide/gpu.ipynb
• site/en/guide/graph_optimization.ipynb
• site/en/guide/intro_to_graphs.ipynb
• site/en/guide/intro_to_modules.ipynb
• site/en/guide/migrate.ipynb
• site/en/guide/mixed_precision.ipynb
• site/en/guide/ragged_tensor.ipynb
• site/en/guide/random_numbers.ipynb
• site/en/guide/saved_model.ipynb
• site/en/guide/tensor.ipynb
• site/en/guide/tf_numpy.ipynb
• site/en/guide/tpu.ipynb
• site/en/guide/upgrade.ipynb
• site/en/guide/variable.ipynb
• site/en/r1/guide/autograph.ipynb
• site/en/r1/guide/distribute_strategy.ipynb
• site/en/r1/guide/eager.ipynb
• site/en/r1/guide/keras.ipynb
• site/en/r1/guide/ragged_tensors.ipynb
• site/en/r1/tutorials/_index.ipynb
• site/en/r1/tutorials/distribute/keras.ipynb
• site/en/r1/tutorials/distribute/tpu_custom_training.ipynb
• site/en/r1/tutorials/distribute/training_loops.ipynb
• site/en/r1/tutorials/eager/automatic_differentiation.ipynb
• site/en/r1/tutorials/eager/custom_layers.ipynb
• site/en/r1/tutorials/eager/custom_training.ipynb
• site/en/r1/tutorials/eager/custom_training_walkthrough.ipynb
• site/en/r1/tutorials/eager/eager_basics.ipynb
• site/en/r1/tutorials/estimators/boosted_trees.ipynb
• site/en/r1/tutorials/estimators/boosted_trees_model_understanding.ipynb
• site/en/r1/tutorials/estimators/cnn.ipynb
• site/en/r1/tutorials/estimators/linear.ipynb
• site/en/r1/tutorials/images/hub_with_keras.ipynb
• site/en/r1/tutorials/images/transfer_learning.ipynb
• site/en/r1/tutorials/keras/basic_classification.ipynb
• site/en/r1/tutorials/keras/basic_regression.ipynb
• site/en/r1/tutorials/keras/basic_text_classification.ipynb
• site/en/r1/tutorials/keras/overfit_and_underfit.ipynb
• site/en/r1/tutorials/keras/save_and_restore_models.ipynb
• site/en/r1/tutorials/load_data/images.ipynb
• site/en/r1/tutorials/load_data/tf_records.ipynb
• site/en/r1/tutorials/non-ml/mandelbrot.ipynb
• site/en/r1/tutorials/non-ml/pdes.ipynb
• site/en/r1/tutorials/representation/unicode.ipynb
• site/en/r1/tutorials/sequences/text_generation.ipynb
• site/en/tutorials/customization/basics.ipynb
• site/en/tutorials/customization/custom_layers.ipynb
• site/en/tutorials/customization/custom_training.ipynb
• site/en/tutorials/customization/custom_training_walkthrough.ipynb
• site/en/tutorials/distribute/custom_training.ipynb
• site/en/tutorials/distribute/input.ipynb
• site/en/tutorials/distribute/keras.ipynb
• site/en/tutorials/distribute/multi_worker_with_estimator.ipynb
• site/en/tutorials/distribute/multi_worker_with_keras.ipynb
• site/en/tutorials/distribute/save_and_load.ipynb
• site/en/tutorials/estimator/boosted_trees.ipynb
• site/en/tutorials/estimator/boosted_trees_model_understanding.ipynb
• site/en/tutorials/estimator/keras_model_to_estimator.ipynb
• site/en/tutorials/estimator/linear.ipynb
• site/en/tutorials/estimator/premade.ipynb
• site/en/tutorials/generative/adversarial_fgsm.ipynb
• site/en/tutorials/generative/autoencoder.ipynb
• site/en/tutorials/generative/cvae.ipynb
• site/en/tutorials/generative/cyclegan.ipynb
• site/en/tutorials/generative/dcgan.ipynb
• site/en/tutorials/generative/deepdream.ipynb
• site/en/tutorials/generative/pix2pix.ipynb
• site/en/tutorials/generative/style_transfer.ipynb
• site/en/tutorials/images/classification.ipynb
• site/en/tutorials/images/cnn.ipynb
• site/en/tutorials/images/data_augmentation.ipynb
• site/en/tutorials/images/segmentation.ipynb
• site/en/tutorials/images/transfer_learning.ipynb
• site/en/tutorials/images/transfer_learning_with_hub.ipynb
• site/en/tutorials/interpretability/integrated_gradients.ipynb
• site/en/tutorials/keras/classification.ipynb
• site/en/tutorials/keras/keras_tuner.ipynb
• site/en/tutorials/keras/overfit_and_underfit.ipynb
• site/en/tutorials/keras/regression.ipynb
• site/en/tutorials/keras/save_and_load.ipynb
• site/en/tutorials/keras/text_classification.ipynb
• site/en/tutorials/keras/text_classification_with_hub.ipynb
• site/en/tutorials/load_data/csv.ipynb
• site/en/tutorials/load_data/images.ipynb
• site/en/tutorials/load_data/numpy.ipynb
• site/en/tutorials/load_data/pandas_dataframe.ipynb
• site/en/tutorials/load_data/text.ipynb
• site/en/tutorials/load_data/tfrecord.ipynb
• site/en/tutorials/load_data/unicode.ipynb
• site/en/tutorials/quickstart/advanced.ipynb
• site/en/tutorials/quickstart/beginner.ipynb
• site/en/tutorials/reinforcement_learning/actor_critic.ipynb
• site/en/tutorials/structured_data/feature_columns.ipynb
• site/en/tutorials/structured_data/imbalanced_data.ipynb
• site/en/tutorials/structured_data/preprocessing_layers.ipynb
• site/en/tutorials/structured_data/time_series.ipynb
• site/en/tutorials/text/image_captioning.ipynb
• site/en/tutorials/text/nmt_with_attention.ipynb
• site/en/tutorials/text/text_classification_rnn.ipynb
• site/en/tutorials/text/text_generation.ipynb
• site/en/tutorials/text/transformer.ipynb
• site/en/tutorials/text/word_embeddings.ipynb
• tools/templates/notebook.ipynb
• tools/templates/subsite/g3doc/tutorials/notebook.ipynb

Rendered notebook diffs available on ReviewNB.com.

Format and style

Use the TensorFlow docs notebook tools to format for consistent source diffs and lint for style:

$ python3 -m pip install -U --user git+https://github.com/tensorflow/docs 
$ python3 -m tensorflow_docs.tools.nbfmt notebook.ipynb 
$ python3 -m tensorflow_docs.tools.nblint --arg=repo:tensorflow/docs notebook.ipynb 

If commits are added to the pull request, synchronize your local branch: git pull origin test

[MarkDaoust]

Friendly ping @yashk2810

[lamberta]

Merge conflict, please rebase

[8bitmp3]

That's a great observation about the zero-centered Normal, thanks for bringing this up.

I have a few style-related suggestions, if you don't mind:

• Change the DCGAN spelling from "Dc-gan" (source: https://arxiv.org/pdf/1511.06434.pdf) (in the code comment)
• Change to weight initializer, instead of model initializers (also in the code comment)
• Add a blank line before the generator method
• Add a space in the params section in RandomNormal() after a comma (I think it's PEP 8: https://www.python.org/dev/peps/pep-0008/)

Suggested change

	"# Random Normal model initializers as per Dc-gan paper\n",
	"RN_initializers=tf.keras.initializers.RandomNormal(mean=0.0,stddev=0.02)\n\n",
	"# Random Normal weight initializer as per the DCGAN paper\n",
	"RN_initializers=tf.keras.initializers.RandomNormal(mean=0.0, stddev=0.02)\n\n",

Also, since it's a weight initializer (singular), maybe you should rename the variable to RN_initializer without the s.

Anyway, these are just recommendations 👍