Added Activation Function and Increased Training Epochs in the CSV Tutorial

[juelg]

This PR adds non-linear ReLU activation functions to the CSV tutorial in order to make the networks non-linear. I have noticed that the activation functions are missing as they are set to None when not explicitly passed and also, as to my knowledge, not added anywhere else. This makes the network linear which is probably not what we want in this context as we don't know whether the dataset is linear. Furthermore, stacking several layers does not make much sense in the linear case as it does not add any more capacity to the network.

Using a non-linear neural network also increases the training performance as shown in the evaluation below. However, the difference is only visible if trained for a little longer, this is why I increased the training epochs to 30.

Training on the abalone dataset for 30 epochs over $n=10$ runs without an activation function leads to the following loss (mean $\pm$ std):
$5.10\pm0.04$

Training with the same conditions but with ReLU as activation function:
$4.90\pm0.03$

The training examples later in the tutorial show a similar performance increase when ReLU is used.

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.

[github-actions]

Preview

Preview and run these notebook edits with Google Colab:

• site/en/tutorials/load_data/csv.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 activation-functions

[MarkDaoust]

Right, thanks!

But I'd prefer to keep 10 epochs, just to keep things quick. The point is just to show that it works, not to get all the way to convergence.