Due

CONTENTS

• Introduction
• Lab Instructions
• What to hand in

1. Conx is one module of the Pyro project. Pyro is an open source project, so you are free to go look at the implementation of Conx at any time. It is available on our system at:

   /usr/lib/python2.4/site-packages/pyrobot/brain/conx.py
   

2. There is a collection of materials available online to help you learn Conx.
   • Start with this brief introduction to Conx.
   • You may want to review the material on the basics of neural nets.
   • Focus more closely on how to build networks in Conx. Do all of the exercises in this section except #2.
   • Try the auto-associative example. This example is very similar to the experiment you'll be reproducing for this lab.
   • You may want to refer to the summary of Conx methods.

3. Using the auto-associative network example as a model, reproduce the compression experiment described on slides 20 and 21 of lecture 12. You'll need to create a network with 8 inputs, 3 hiddens, and 8 outputs. The goal of this network is to output the same pattern that appears in the input. Test various network parameter settings.
   • Once you have settled on some reasonable settings, run the experiment three times and describe the results.
   • Does each network come up with the same hidden layer representations?
   • Do the hidden layer representations learn something similar to binary numbers?
   • Can this task be learned with only 2 hidden nodes? Try this and describe the results.

4. Use the picture data to train a neural network to determine whether the person in the picture is wearing sunglasses.
   • Download the training data which contains 624 images stored in PGM format. To uncompress first run:

     tar xvfz pics.tar.gz
     

     It will create a directory called pics with all the pictures in it. You can view one of these images using the xv command. Each file is named according to the following convention:

     
     userid_pose_expression_eyes_scale.pgm
     

     • userid is the user id of the person in the image. This field has 20 values: ani2, at33, boland, bpm, ch4f, cheyer, choon, danieln, glickman, karyadi, kawamura, kk49, megak, mitchell, night, phoebe, saavik, steffi, sz24, and tammo.
     • pose is the head position of the person, and this field has 4 values: straight, left, right, and up.
     • expression is the facial expression of the person, and this field has 4 values: neutral, happy, sad, and angry.
     • eyes: is the eye state of the person, and this field has 2 values: open and sunglasses.
     • scale: is the scale of the image. All of the images are of type 4 which indicates a quarter-resolution image (32 x 30).

   • Here is a text file that contains a list of all the pictures and this is a program that will randomly divide the filenames into equal size subsets for training and validation
   • There are some python functions to help you process these image files and also to help you visualize the weights as an image such as the ones at the end of Lecture 12.
   • When designing your experiment be sure to use a training set and a validation set so you can report an error rate and to avoid overfitting Conx provides methods for setting up a cross validation corpus and using it. Look in implementation details. You will be able to use a loop to create the corpus using code like the following:
     

     length = # of test case
     corpus = []
     for i in range(0,length):
             corpus.append({"input" : in[i], "output" : out[i]})
     

   • Through trial and error determine the best network architecture to accomplish the task. We'd like the smallest hidden layer size for which the task can be consistently learned.
   • Once you have trained a neural network in Conx, you can save the hidden layer activations as follows:

     n.setLearning(0)            #turn learning off
     n.setOrderedInputs(1)       #turn off the default random ordering for inputs
     hid = n.getLayer('hidden')  #get a pointer to the hidden layer
     hid.setLog("hiddens")       #set a log file for the hidden layer
     n.sweep()                   #sweep through the inputs in order
     hid.closeLog()              #close the log file for the hidden layer
     

   • In your write up describe the network that you used, describe the hidden layer weights that were found, and report the error rate of the network