1-1: Introduction Prof. J.C. Kao, UCLA. Introduction. Course information Syllabus Other logistics

Transcription

1 1-1: Introduction Prof. J.C. Kao, UCLA Introduction Course information Syllabus Other logistics

2 1-2: Introduction Prof. J.C. Kao, UCLA Course information Course: Instructor: Teaching assistants: ECE 239AS - Neural networks and deep learning Jonathan Kao kao@seas.ucla.edu Cheng Zheng superzhengcheng@ucla.edu Tianwei Xing twxing@ucla.edu Grading: 40% homework, 40% midterm, 19% final project, 1% participation Please do not us individually for class questions. We will not respond to individual s with class questions. All class-related questions should be sent to the class account: ece239as.w18@gmail.com

3 1-3: Introduction Prof. J.C. Kao, UCLA Course information, cont. Prof. Kao office hours Office: Eng IV H Hours: Th 11:15am - 1:00 pm Cheng office hours Office: Eng IV Hours: Tu 10:00am - 12:00 pm Tianwei office hours Office: Eng IV Hours: Tu 3:00pm - 5:00 pm Sections: Fri 12-12:50pm, Lakretz 120 w/ Cheng Fri 2-2:50pm, Bunche 3170 w/ Cheng Fri 3-3:50pm, MS 5128 w/ Tianwei Fri 4-4:50pm, MS 5127 w/ Tianwei

4 1-4: Introduction Prof. J.C. Kao, UCLA Course information, cont. Textbook: Deep Learning Ian Goodfellow, Yoshua Bengio, and Aaron Courville Available here: We will use CCLE for course materials. Course website:

5 1-5: Introduction Prof. J.C. Kao, UCLA Rough syllabus Date (2018) Lecture Content 08 Jan 1 Overview to deep learning 10 Jan 2 Introduction to machine learning I 15 Jan - MLK Holiday HW #1 released, due 22 Jan 17 Jan 3 Introduction to machine learning II 22 Jan 4 Supervised classification & gradient descent principles HW #2 released, due 29 Jan 24 Jan 5 Fully connected neural networks 29 Jan 6 Backpropagation HW #3 release, due 05 Feb 31 Jan 7 Regularizations for training neural networks 05 Feb 8 Optimization for training neural networks HW #4 released, due 12 Feb 07 Feb 9 Convolutional Neural Networks I 12 Feb 10 Convolutional Neural Networks II HW #5 released, due 19 Feb 14 Feb 11 Recurrent neural networks I 19 Feb - Presidents Day Holiday 21 Feb M Midterm, in class Project released, due 16 Mar 26 Feb 12 Recurrent neural networks II 28 Feb 13 Variational autoencoders I 05 Mar 14 Variational autoencoders II 07 Mar 15 Other topics in deep learning 12 Mar 16 Other topics in deep learning II 14 Mar 17 Overview

6 1-6: Introduction Prof. J.C. Kao, UCLA Pre-requisites This class requires a solid understanding in probability and linear algebra. It also requires coding experience. The class will be taught entirely in Python. If you have only had exposure to MATLAB there may be some ramp up time to familiarize yourself with Python. Pre-requisite topics I will assume you know. Probability: independence, conditional probability, Bayes rule, multivariate Gaussian distribution, marginalization, expectation, variance Linear algebra: basic matrix operations, span, rank, range and null space, eigenvalue decomposition, singular value decomposition, pseudoinverse

7 1-7: Introduction Prof. J.C. Kao, UCLA Academic integrity UCLA embraces the core values of integrity, excellence, accountability, respect, and service through the True Bruin program I take academic integrity very seriously; students caught cheating or violating these principles will face disciplinary action. Please refer to the UCLA student conduct code: In this class, unacceptable behavior includes plagiarizing the work of others, plagiarizing code, and copying another person s exam. In accordance with UCLA policy, any instance of suspected academic dishonesty will be immediately reported to the Dean of Students Office and zero credit will be given for any work determined to be dishonest.

8 1-8: Introduction Prof. J.C. Kao, UCLA A few notes on this class This class is being taught for the first time in the Winter quarter of AY at UCLA. These notes are being used for the first time, and will likely have errors. Please report them via . I will update the lecture notes appropriately. The syllabus is rough. Some topics may take longer than expected. Other topics may take shorter than expected. As a result, some topics may not be covered; alternatively if there is time at the end, additional topics may be covered. Similarly, the assignments for this class this first time around will also be relatively rough. While we will do our best to make the assignments as error-free as possible, we do recognize that inevitably, the first time around, we likely won t get everything right. We apologize in advance for any inconveniences as a result of this. We will do our best to correct any errors in assignments as quickly as possible.