COMP 193A/293A Summer I 2024 Syllabus

This course aims at introducing deep neural networks as an effective approach in constructing machine learning-based solutions. The course helps students in the design and development of neural networks in a given application domain, using hands-on approach. Topics include tensors, learning basics, artificial neural networks, and convolutional neural networks among other topics.

Prerequisites: COMP 053 and ECPE 170 with a "C-" or better.
When enrolling in this course, you should be relatively proficient in using Python. In addition to being a competent Python programmer, you should have an understanding of partial derivatives and gradients, probability, and using common data structures, e.g., lists, arrays, maps, classes, etc. We will use PyTorch to develop models and train them. Beforehand knowledge of PyTorch is not required.

Administration

Instructor: Sepehr Amir-Mohammadian
Email:

Class time/location: Class is held in asynchronous online mode

Teaching Assistant:
Email:
Office hours:

Learning Objectives

The vision for this course is: What do I, as a computer scientist, need to understand the structure of deep neural networks and what are the different features by which I can customize these networks in a given project?

Labs
Quizzes
Class discussions, and reading/watching assignments

Use tensors and functions on tensors to facilitiete the data being fed to deep neural networks.
Transform data of different types to tensors.
Understand and program the basics of training a model: the general process, loss and its minimization, backpropagation, optimization strategies, overfit models, training vs. validation, model accuracy
Understand and program artificial neural networks: artificial neurons, linear transformation, multi-layer networks, and activation functions
Use well-known datasets, and prepare data for training purposes, e.g., data normalization
Implement and train fully-connected layers in neural networks for classification problems, e.g., in image recognition
Understand the limitation of fully-connected networks and the necessity of using convolutional neural networks in certain domains
Understand and program convolutional neural networks: kernels, convolutions, padding, downsampling, etc.
Train convolutional neural networks: extending the capacity of the networks, combating overfitting in the networks

University of the Pacific Core Competencies: This course reflects the following university-wide core competenceies in the undergraduate program:

Critical Thinking
Information Literacy
Quantitative Reasoning

Outcomes for COMP program: The assessment plan for this course comprises the following outcomes identified by ABET:

Outcome 1: "Analyze a complex computing problem and to apply principles of computing and other relevant disciplines to identify solutions."
Outcome 2: "Design, implement, and evaluate a computing-based solution to meet a given set of computing requirements in the context of the program’s discipline."
Outcome 6: "Apply computer science theory and software development fundamentals to produce computing-based solutions."

Collection of Work for Assessment: Student work may be retained to assess how course learning objectives are being met and for accreditation purposes.

Course Material

We will use the following textbook along with additional resources that are referred through the semester.

Deep Learning with PyTorch by Eli Stevens, Luca Antiga, and Thomas Viehmann, Manning Publications Co., 2020.

The slides, lectures, assignments, supplementary material, etc. will be provided through Canvas LMS.

Tensors and their use in machine learning
Transformation of data to tensors
Basics of learning
Artificial neural networks
Convolutional neural networks

Grading and Attendance Policy

Grades for the course are assigned on the scale below:

A	A-	B+	B	B-	C+	C	C-	D+	D	F
[93,100]	[90,93)	[87,90)	[83,87)	[80,83)	[77,80)	[73,77)	[70,73)	[67,70)	[60,67)	[0,60)

Final grades will be assigned based on several performance factors. These factors and their quantitative contribution to the final grade are as follows:

Reading assignments: 0%
Quizzes: 30%
Labs: 70%

Attendance: This class is held online asynchronously. So, there are no attendance requirements.

Assignment Guidelines

Reading/Watching Assignments: These assignments will be released on the course Canvas page at most one or two days before each lecture session. Students must study/watch the referred material before each class session. There are not any submissions for reading/watching assignments, however it is highly recommended to follow the reading assignments before watching the lecture videos. This would help you better prepare for quizzes and labs. Reading/Watching assignments are considered as individual efforts.
Quizzes: Each session comes with a brief quizz (10-15 minutes) that reflects on the material covered in the reading/watching assignment associated with that class session. Quizzes are published and automatically graded in Canvas. There will be no extensions to quiz due dates. These assignments are considered as individual efforts.
Labs: Labs will be released on the course Canvas page. The due date for labs might be postponed, if considered necessary during the semester. Lab assignments are considered to be group-based.

Late submission policy: Delliverables for quizzes and labs will not be accepted after the due date.
All assignments will be considered individual efforts unless otherwise specified, and will be treated as such under the Academic Honesty Policy.

Academic Honesty

The Honor Code at the University of the Pacific calls upon each student to exhibit a high degree of maturity, responsibility, and personal integrity. Students are expected to:

Act honestly in all matters
Actively encourage academic integrity
Discourage any form of cheating or dishonesty by others
Inform the instructor and appropriate university administrator if she or he has a reasonable and good faith belief and substantial evidence that a violation of the Academic Honesty Policy has occurred.

Violations will be referred to and investigated by the Office of Student Conduct and Community Standards. If a student is found responsible, it will be documented as part of her or his permanent academic record. A student may receive a range of penalties, including failure of an assignment, failure of the course, suspension, or dismissal from the University. The Academic Honesty Policy is located in Tiger Lore and online.

Course-specific Honor Code Policy: Engineering is generally a cooperative endeavor and collaborative learning can be a valuable experience for all involved. However, proper assessment (i.e., grading) requires that work be done by individuals. To balance these two requirements, the following policy will apply:

Collaborative work on course assignments is encouraged. This includes working together on planning solution strategies and helping each other to debug programs.
Collaboration must stop short of someone else writing your assignment. You may not directly copy the work of another student. You also may not copy the work of another student, and then modify it so that it does not look the same as the original author's work. It is your responsibility to ensure that the work you submit is an honest representation of your own understanding of the material.

Marginal cases will be resolved by oral examination of the student(s) involved. If they each understand the material in the assignment, it will be considered honest collaboration. If they do not, then it will be considered academic dishonesty.

In many cases, it may be possible to identify reusable source code from textbooks, web sites or other resources that can help you with assignments. You are permitted to use such references provided that:

The amount of code reused does not exceed 25% of the total assignment length, and
In the source code comments, you clearly identify any code that you did not write, state where it came from, and to what extent you modified it.

You are responsible for understanding the theory behind all algorithms or source code used, regardless of their source.

Accommodations for Students with Disabilities

If you are a student with a disability who requires accommodations, please visit pacific.edu/disabilities to contact the Office of Services for Students with Disabilities (SSD) for information on how to request accommodations while at Pacific.

Students who have not previously registered for accommodations can request services by visiting pacific.edu/disabilities and selecting New Students Apply Here. Once registered, students will be asked to provide documentation of their disability, and meet with the accommodation specialists to determine reasonable accommodations.
Students who have previously been approved for services with SSD can request accommodation(s) letters each semester by selecting the Returning Students Login Here link located on pacific.edu/disabilities

The Office of Services for Students with Disabilities is located in the McCaffrey Center, Second Floor. Phone: 209-946-3221. Email: ssd@pacific.edu. Online: pacific.edu/disabilities

Nondiscrimination Policy

The University of the Pacific does not discriminate in the administration of any of its educational programs, admissions, scholarships, loans, athletics, or other University activities or programs on the basis of race, color, national and ethnic origin, handicap, sexual orientation or preference, sex or age.

COMP 193A/293A: Introduction to Deep Learning

Course Description