Neural Networks - Fall 2020 - 27-436

Staff:

Prof Gal Chechik (webpage)
Zach Cohen

Time and place:

Place: Link to Zoom by invitation only for registered students
Time: Mondays 4PM -- 5:30PM

Administration:

There are 8 home assignments in the course, and a final project. Home assignments are a mix of questions and programming assignments. If a home assignment is submitted late, it will be penalized 5 points for each day. Programming assignments can be submitted in pairs. Form-based assignments should be submitted individually. Pen-end-paper assignments can be discussed in pairs, but should be submitted individually. Every student should write their own solution, submit it under their name, and note the name of another student that they discussed the material with.

Based on University guidelines, parents to small children can chose to be exempt from submitting the last two home assignments. If they decide to do so, their grade will be based on Home assignments 1-6. Students that want to take this option should submit an empty Assignment 7 and Assignment 8 and declare that they are parents of small children. Submission of these assignments should be done by the deadline. .

There were also two class assignments, which can get students gain up to 6.4 bonus points together, if their assignment average is above 60. The bonus will be computed as follows: max(A, A*0.94 + B1*0.05 + B2*0.05), where A is the average grade of home assignments and B1 is the grade in the first bonus assignment, and B2 is the grade in the second bonus assignment.

Originally, there was a mismatch between the text in the syllabus and the text here. The corrected weights are as follows: The final grade in the course is computed based on 30% of home assignments and 70% of final project. Students must submit all home assignments to deserve credit in the course.

Home assignments:


	Link	Submission date
	Ex 1 - Generalization	2020-11-01
	Ex 2 - Basic Perceptron	2020-11-09
	Ex 3 - Regularize the Perceptron	2020-11-21
	Ex 4 - Max Likelihood, Logistic regression, SoftMax	2020-11-30
	Ex 5 - Multi layer perceptron	2020-12-10
	Ex 6 - Convolutional neural networks	2020-01-03
	Ex 7 - Review 1	2020-01-10
	Ex 8 - Review 2	2020-01-17
	There is no Ex 9
	Final project Instructions , Data	2020-04-05

List of classes:


	Date	Topic	Zoom recording	online material
1	2020-10-19	Introduction, Generalization, Polynomial regerssion	Recording	Slides
2	2020-10-26	Linear classifiers, Optimiztion with SGD	Recording	Slides
3	2020-11-02	Linear classifiers and noise. Logistic regression	Recording	Slides
4	2020-11-09	Non-linear classification. Multiclass classifiction	Recording	Slides
5	2020-11-16	Multi-layer perceptrons	Recording	Slides
6	2020-11-23	Deep networks, Convolutional neural networks	Pre-recorded , Discussion	Slides
7	2020-11-30	Overview of deep architectures	Recording
8	2020-12-07	Handling discerete data with word embeddings, Cross entropy	Recording	Slides
9	2020-12-14	Hanuka. Self-reading	Slides 1 Slides 2
10	2020-12-21	Visualization part 1	Recording	Slides
11	2020-12-28	Visualization part 2		Slides
12	2020-01-04	Review of class assignment	Recording
13	2020-01-11	Probabilistic neurons	Recording	Slides
14	2020-01-18	No class

Related material and courses

Pattern Recognition and Machine Learning , Chris Bishop
Deep Learning Ian Goodfellow and Yoshua Bengio and Aaron Courville, MIT Press, 2016.