BYU CS classes

To read current papers on DNN research and translate them into working models. To experiment with DNN-style regularization methods, including Dropout, Dropconnect, and L1/L2 weight regularization.

For this lab, you will need to implement three different regularization methods from the literature, and explore the parameters of each.

1. You must implement dropout (NOT using the pre-defined Tensorflow layers)
2. You must implement dropconnect
3. You must experiment with L1/L2 weight regularization

You should turn in an iPython notebook that shows three plots, one for each of the regularization methods.

1. For dropout: a plot showing training / test performance as a function of the “keep probability”.
2. For dropconnect: the same
3. For L1/L2: a plot showing training / test performance as a function of the regularization strength, \lambda

An example of my training/test performance is shown at the right.

Your notebook will be graded on the following:

• 40% Correct implementation of Dropout
• 30% Correct implementation of Dropconnect
• 20% Correct implementation of L1/L2 regularization
• 10% Tidy and legible plots

This lab is a chance for you to start reading the literature on deep neural networks, and understand how to replicate methods from the literature. You will implement 4 different regularization methods, and will benchmark each one.

To help ensure that everyone is starting off on the same footing, you should download the following scaffold code:

For all 4 methods, we will run on a single, deterministic batch of the first 1000 images from the MNIST dataset. This will help us to overfit, and will hopefully be small enough not to tax your computers too much.

Part 1: implement dropout

For the first part of the lab, you should implement dropout. The paper upon which you should base your implementation is found at:

Dropout

The relevant equations are found in section 4 (pg 1933). You may also refer to the slides.

There are several notes to help you with this part:

1. First, you should run the provided code as-is. It will overfit on the first 1000 images (how do you know this?). Record the test and training accuracy; this will be the “baseline” line in your plot.
2. Second, you should add dropout to each of the h1, h2, and h3 layers.
3. You must consider carefully how to use tensorflow to implement dropout.
4. Remember that when you test images (or when you compute training set accuracy), you must scale activations by the keep_probability, as discussed in class and in the paper.
5. You should use the Adam optimizer, and optimize for 150 steps.

Note that although we are training on only the first 1000 images, we are testing on the entire 10,000 image test set.

In order to generate the final plot, you will need to scan across multiple values of the keep_probability. You may wish to refactor the provided code in order to make this easier. You should test at least the values [ 0.1, 0.25, 0.5, 0.75, 1.0 ].

Once you understand dropout, implementing it is not hard; you should only have to add ~10 lines of code.

Part 2: implement dropconnect

The specifications for this part are similar to part 1. Once you have implemented Dropout, it should be very easy to modify your code to perform dropconnect. The paper upon which you should base your implementation is

The dropconnect paper

Important note: the dropconnect paper has a somewhat more sophisticated inference method (that is, the method used at test time). We will not use that method. Instead, we will use the same inference approximation used by the Dropout paper – we will simply scale things by the keep_probability.

You should scan across the same values of keep_probability, and you should generate the same plot.

Dropconnect seems to want more training steps than dropout, so you should run the optimizer for 1500 iterations.

Part 3: implement L1/L2 regularization

For this part, you should implement both L1 and L2 regularization on the weights. This will change your computation graph a bit, and specifically will change your cost function – instead of optimizing just cross_entropy, you should optimize cross_entropy + lam*regularizers, where lam is the \lambda regularization parameter from the slides. You should regularize all of the weights and biases (six variables in total).

You should create a plot of test/training performance as you scan across values of lambda. You should test at least [0.1, 0.01, 0.001].

Note: unlike the dropout/dropconnect regularizers, you will probably not be able to improve test time performance!

To generate a random binary matrix, you can use np.random.rand to generate a matrix of random values between 0 and 1, and then only keep those above a certain threshold.