Assignment - 3
The following papers are published in the early stage of machine learning, but they are
excellent papers to review. Read all three papers. Choose one paper and summarize it in your
own wording. You can download these papers from Module 5-Review papers on Western Online.
Do not have more than 3 pages for the summary.
1. Parallel Neural Networks that Learn to Pronounce English Text by Terrence J. Sejnowski and
Charles R. Rosenberg
2. Efficient Training of Artificial Neural Networks for Autonomous Navigation by D.A.
Pomerlau
3. SexNet: A Neural Network Identifies Sex from Human Faces by B.A. Golomb, D.T.
Lawrence, and T.J. Sejnowski
Review a paper by the following format.
1. Problem definition: What is the problem?
2. Data set: Describe the input and output data.
3. Idea to solve the problem
4. System architecture: Input, Hidden, Output layers
5. Experimental results
2)An excellent overview paper on Deep Learning was published in Nature. The authors, Yann LeCun, Yoshua Bengio, and Geoffrey Hinton, are pioneers and leading scientists in Deep Learning field.You can access the paper through http://www.nature.com/nature/journal/v521/n7553/full/nature14539.html or download it from Module 5-Review papers on Western Online.
· Yann LeCun, Yoshua Bengio, and Geoffrey Hinton. “Deep learning.”, Nature 521.7553, 436-444. (2015)
You can skip sections of “Distributed representations and language processing” and “Recurrent neural networks”.
Write three pages of summary on this review paper. Do not include pictures in your submission.
Do not just write down the contents of the paper. Read the paper and submit the assignment with your own writing. Try to focus on some interesting points, for example:
What kind of algorithms are used? How is Backpropagation used?
What kind of applications are improved in Deep Learning? Why CNN is so successful?
What is representation learning? And why is it important? How is generalization explained in this paper?
What kind of animals are introduced in the paper and why? Why Deep Learning will be successful in the future?