605.746.81 - Advanced Machine Learning

Expanded Course Description

Prerequisites: 605.649 Introduction to Machine Learning; multivariate calculus, linear algebra, probability and statistics, discrete mathematics.

Course Structure

Details on the course structure can be found in the Course Outline. Each course module runs for a period of seven (7) days. Due dates for readings and other assignments are referred to by the day of the module week in which they are due. For example, if a reading assignment is to be completed by Day 3 and the module started on Monday, then the reading assignment should be completed by Wednesday or the 3rd day of the module. Please refer to the Course Outline for the specific start and end dates for each module in this course. Given the slightly odd schedule for the modules, please pay careful attention to the course calendar.

Course Topics

• Introduction
• Experimental Methods
• Computational Learning Theory
• Nonparametric Learning
• Bayesian Learning
• Artificial Neural Networks
• Reinforcement Learning
• Evolutionary Computation
• Inductive Logic Programming
• Data/Dimensionality Reduction
• Kernel Methods
• Ensemble Learning

Course Goals

This is a research-based course. Therefore, the primary goals of this course are to develop broad understanding of the issues in developing and analyzing modern machine learning methods, and to develop a deeper understanding of at least one specific machine learning topic through an in-depth research project.

Course Learning Outcomes (CLOs)

• Formulate and assess problems in machine learning and data mining.
• Assess the strengths and weaknesses of several machine learning algorithms.
• Assess and understand the key commonalities and differences in applications of machine learning to agent control and data mining.
• Apply techniques in machine learning to problems in agent control or data mining.
• Develop a proposal for an extended research project.
• Design and conduct an experiment in machine learning.
• Prepare a written, technical paper on individual research.
• Deliver a presentation, describing the results of individual research.

Textbooks

Required

There is no required textbook for this course; however, students will be required to read several recent research papers in machine learning. The required paper in each module is identified in the Assignments area of that module. These articles can be accessed through the Sheridan Libraries proxy server to the identified journals.

Detailed course notes are provided to all students in the Course Information area as well as in the respective modules.

Each student will also be required to read and critique a PhD dissertation. A list of acceptable dissertations can be found on the Machine Learning PhD Dissertations page, available from the Course Information section of this course. If students need help accessing their desired dissertation, a copy can be obtained from the instructor.

Optional

Mitchell, T. M. (1997). Machine learning. New York, NY: McGraw-Hill. ISBN-10: 0070428077 ISBN- 13: 9780070428072.

Alpaydin, E. (2020). Introduction to machine learning (4rd ed.). Cambridge, MA: The MIT Press. ISBN 9780262043793.

Bishop, C. M. (2006). Pattern recognition and machine learning. New York, NY: Springer, ISBN-10: 0387310738 ISBN-13:9780387310732.

Other Materials & Online Resources

A headset or microphone and speakers are required for this course. It is highly recommended that you purchase an inexpensive USB headset with microphone for your computer to optimize your participation in online sessions. USB headsets can be found online or in electronics stores for approximately $20–$30. A microphone and speakers built into your computer or a separate microphone and computer speakers are the minimum requirement.

A webcam is also required for this course to be used when delivering your project presentations and when participating in office hours.

Student Coursework Requirements

Assignment Guidelines and Expectations

As a blanket restriction, since the intent of the assignments in this courses is to foster the development of research skills in the area of machine learning, and since a significant component of developing such skills is in expressing one's understanding about technical subjects, the use of generative AI-based tools such as Gemini or ChatGPT are strictly prohibited in the completion of any of the course assignments, including generating discussion questions, participating in a discussion, writing paper summaries or dissertation critiques, or in the completion of the research proposal and project report.

Discussion Leadership

This course is formatted as a seminar in which research papers are read and discussed each week. To make the course more interesting and to encourage involvement by all students in the discussion, the seminar will be conducted such that the students will be responsible for presenting the weekly material.

A unit will be structured as follows. Each week, students will be required to read the topical content as well as the assigned paper. Both should be completed by Day Three of the current week. Starting that day, the assigned discussion leader will present an overview of the paper(s) for the week and formulate questions (if one leader, five questions; if two leaders, three questions each) and issues for class discussion. The overview and each question shall be posted in separate threads in the Discussions area appropriate for the week.

If more than one student is assigned to lead, then each student must submit a separate overview of the paper. The overview should include a review of the paper, a review of any related work, either identified in the paper or arising from a personal literature review, and a review of the issues and significance of the work reported. The overview should be no more than five pages long, single-spaced. The evaluation criteria for paper summaries are as follows:

• Summary of paper (25%)
• Identification of key issues (25%)
• Discussion of the claimed contributions (25%)
• Review of related material, including references (15%)
• Construction and readability (10%)

Reminder: The use of generative AI-based tools such as Gemini or ChatGBT is not permitted when writing the discussion summary.

After the overview has been presented, the class will be called upon to engage in discussion of the questions and issues raised by the discussion leader. Each student leader must also act as facilitator for all of the discussion threads whether corresponding to their question or not. The instructor will participate as another member of the discussion, interjecting additional material as necessary to provide information on background and current research in the field. The instructor may also play a “devil’s advocate” role in order to provoke a reaction and further discussion. Student leaders should not take on this role.

To prepare for leading discussion, the leader shall read the paper very carefully, being sensitive to issues such as:

• Algorithms proposed
• Technical approach used in experiments and analyses
• Comparisons with other methods
• Claims of contributions made by the paper
• Biases of the investigators (either explicit or implicit)
• Deficiencies in the paper or the research reported
• Directions for future research
• Possible ideas for course research projects

The evaluation criteria for discussion leadership are as follows

• Ability to pose properly formed, open-ended questions for discussion (30%)
• Ability to facilitate discussion continually throughout module (30%)
• Ability to demonstrate understanding of paper content, as evidenced by discussion (20%)
• Daily participation in discussion all discussion forums for paper, Day 3 through Day 7 (20%)

Dissertation Critique

Fundamentally, this is a research-oriented course, and a large number of topics will be covered as a foundation for a researcher to apply in solving complex problems involving machine learning algorithms. Furthermore, this course is oriented around introducing the student to current research in the field of machine learning. Unfortunately, in a course such as this, it is difficult during a given week to explore any one topic in depth. Therefore, each student will be responsible for selecting a PhD dissertation from a list of several provided by the instructor and writing a critique of the research reported in that dissertation. Most of these dissertations are available through UMI ProQuest, which is accessible through the JHU Proxy server. Any that are difficult to find can be requested from the instructor.

The critique should include a summary of the research reported, a discussion of the major contributions claimed, and an assessment of the significance of those contributions and of the research itself. The critique should also include a brief literature review of the topic related to the thesis, discussion of relevant algorithms, and application areas for the research reported. Where appropriate, the critique should include a comparison with other issues discussed in class. When feasible, students are encouraged to select a dissertation that is related to their course projects.

The evaluation criteria for the critique are as follows:

• Overview of the research reported (20%)
• Review of the related literature, including references (15%)
• Major contributions of the thesis (20%)
• Understanding of techniques and algorithms (20%)
• Application areas (15%)
• Proper construction and readability of paper (10%)

The critique is limited to be no more than ten (10) pages (single spaced), including all figures, tables, and references. Critiques will be formatted using either the JAIR format (http://www.jair.org) or the JMLR format (http://www.jmlr.org). Note that these formats are essentially identical. Papers longer than 10 pages will be truncated down to the first 10 pages and graded as if subsequent material was not included.

Reminder: The use of generative AI-based tools in the writing of your dissertation critique is not permitted.

Research Projects

As a semester long project, each student in this course will be responsible for completing an independent research project in machine learning. This project will provide direct experience in proposing and executing a complete research project over the length of the course. The project can be experimental or theoretical. If an experimental project is proposed, be prepared to include enough theoretical work to explain or motivate the work. If the project is theoretical, some experimentation should be included to demonstrate whatever results are obtained. Although an application-oriented project is permitted, the focus of your project is required to be on advancing the algorithms and not the application. 

Each individual will prepare a short proposal describing the intended research. This proposal must be approved by the instructor prior to commencing the major portions of the research. Obviously, some amount of research should be done to prepare the proposal. The proposal should include a brief literature survey on the topic area, a clear statement of the problem to be solved (including a clear, testable hypothesis), and a description of the approach to be taken. All proposals will be formatted using either the JAIR (http://www.jair.org) or JMLR (http://www.jmlr.org) paper format. The proposal shall be no more than eight (8) pages in length using this format, including all tables, figures, and references. The proposal is due around the time typical for a midterm exam.

The evaluation criteria for proposals are as follows:

• Statement of problem and hypothesis (25%)
• Proposed approach (25%)
• Relevance and motivation for the topic (25%)
• Literature review, including references (15%)
• Construction and readability of proposal (10%)

Reminder: The use of generative AI-based tools in the writing of your project proposal is not permitted.

The actual execution of the project is left entirely at the discretion of the individual. Any computer and programming language may be used to support the project, and additional tools for analysis and presentation (e.g., MATLAB and Excel) are encouraged. At the end of the project, each individual will be required to submit a comprehensive research report. This report will include background and discussion of previous work done related to the topic, a clear description of the problem to be solved, discussion of the approach taken, in depth discussion of any algorithms used or developed, detailed presentation of the results obtained, discussion of the importance and implications of the results, directions for future work, and references. The author(s) should use the research papers read in this course as guidance for what research papers look like. Note that submitting code is not required.

All research reports are required to be prepared using the JAIR (http://www.jair.org) or JMLR (http://www.jmlr.org) format and shall be no more than fifteen (15) pages long (including all figures, tables, and references). Papers longer than 15 pages will be truncated down to the first 15 pages and graded as if subsequent material was not included.

The evaluation criteria for the report are as follows:

• Statement of problem and hypothesis (15%)
• Approach (20%)
• Background and related work in field, including references (15%)
• Results (15%)
• Discussion and analysis of results (20%)
• Relevant future work (5%)
• Construction and readability of report (10%)

Reminder: The use of generative AI-based tools in the writing of your project report is not permitted.

During the final weeks of the class, results of research projects will be presented in class. Each individual will give a short presentation of the research performed and the results achieved.

The format for the presentation will be similar to a research talk given at a conference, except that we will be using online tools (e.g., Adobe Presenter or Adobe Connect) to deliver the presentation. Presentations will give the class a chance to see other projects and to provide feedback to the student. Note that the student is required to prepare visual aids (e.g., PowerPoint presentations) for their talks. Presentations will be scheduled to last no more than one hour where the actual presentation will be no longer than 20 minutes (uninterrupted), followed by 30–40 minutes of question-and-answer discussion.

The evaluation criteria for the presentation are as follows:

• Statement of problem and hypothesis (15%)
• Approach (20%)
• Background and related work in field (references should be in paper) (15%)
• Results (15%)
• Discussion and analysis of results (20%)
• Relevant future work (5%)
• Clarity in presentation (10%)

Participation Grading Criteria

As a seminar, online discussion is a key component of this course. Each week, at least one student will be assigned as a discussion leader. Since discussion leadership is graded separately, that student's initial discussion questions will not count toward the class participation grade that week. For the discussion, questions and comments will be posed, either by the discussion leader or the professor.

Discussions will be weighed equally in the final grade. That grade will be based on the following approach. Students are required to make substantive contributions to all discussions over multiple days. Thus, each student is expected to post at least two substantive posts in response to each question, and these two posts must occur on different days. Substantive posts either introduce new ideas to the discussion or provide a significant elaboration on a prior post. Simply responding, "I agree" or similar will not count. Furthermore, each student is expected to contribute posts over at least three days for the entire discussion. Given this requirement, participation grades will be calculated as follows:

40 x % questions with substantive posts +

30 x % questions with multiple posts on multiple days +

min {30, 10 x #days posted}

For example, if during Module 4, five questions are posed and student responds to four, has multiple responses provided over two days on two of them, and that student only posts over two days, the grade for that week would be 40 x 4/5 + 30 x 2/5 + 10 x 2 = 64/100. As another example, if during Module 8, six questions are posed and the student respond to all six, has multiple posts over two days on four, but that student posts over four days, the grade for that week would be 40 x 6/6 + 30 x 4/6 + min{30, 10 x 4} = 90.

Reminder: The use of generative AI-based tools in the writing of your discussion posts is not permitted.

Guidelines and Requirements for Substantive Posting

This course is structured in a way to incorporate in-depth seminar-like discussions of recently published research papers in the field. Successful discussions involve a mix of quantity and quality in the posts. The minimal quantity requirements are specified elsewhere; here we focus on what constitutes substantive posting in a seminar discussion (and what does not).

Ultimately, the extent to which a post extends a discussion is the primary criterion for determining if it is substantive. Here are properties of a substantive post.

• Posts that target the discussion question or a previous substantive post and extend the discussion in a useful direction.
• Posts that provide an in-depth analysis of a point from the paper or a previous post.
• Posts that draw in external resources and explain their relevance to the discussion.
• Posts that demonstrate synthesis of ideas, by providing additional analysis, empirical testing, experimentation, etc.
• Posts that hypothesize future directions, combined with discussing a strategy for testing that hypothesis.

Here are properties of non-substantive posts that are still useful.

• Posts that simply ask questions, either about the paper or about what someone else posted.
• Posts that are “one liners” that still have an interesting point but lack explanation.
• Posts that answer someone else’s question by pointing to a discussion in the paper or online without elaboration.
• Posts that introduce a nugget of an interesting, new idea without developing that idea.

Here are properties of non-substantive posts that either do not add to the discussion or detract from the discussion.

• Posts that in words or essence simply say “I agree” or “thank you.”
• Posts that simply repeat the content from the paper or from another post without elaborating, extending, or giving evidence of synthesis.
• Posts that reflect the poster has not read preceding posts in the discussion, thus either leading to a repeat of content or a misstatement that was already addressed.
• Posts that are completely off topic.
• Posts that link to other resources without explaining them.

For grading purposes, the emphasis is being placed on the substantive posts while still recognizing that some non-substantive posts are still useful in advancing the discussion. These latter posts are appreciated but are not counted in the grade.

In general, discussion posts should be regarded as mini-essays and be written with the appropriate essay structure in mind. As a rule of thumb, the starting point for such a substantive post is one that presents a main idea (topic sentence) followed by several statements supporting that idea (generally 3-5). Multiple paragraphs may be necessary to make the point. One or two sentence posts are rarely substantive. Note that taking a multi-paragraph post and breaking it into multiple posts will only receive credit as if it was a single post.

Grading Policy

Grading is based on online asynchronous discussions, discussion leadership, ability to report on progress in the field through oral presentation and written critique, and the ability of the student to design and implement a research project. Students will be responsible for periodically preparing a summary of one of the research papers for use by the class and then leading class discussion on the paper. Each student will also conduct a research project, documented with a formal, technical paper describing the experimental method and results. Final grades will be determined by the following weighting:

A grade of A indicates achievement of consistent excellence and distinction throughout the course—that is, conspicuous excellence in all aspects of assignments and discussion in every week. A grade of B indicates work that meets all course requirements on a level appropriate for graduate academic work. 

EP uses a +/- grading system (see “Grading System”, Graduate Programs catalog, p. 10).

Course Policies

Late Submission Policy

Being that we are all working professionals, and time management is of critical importance, the purpose of this document is to lay out the course policy with respect to completing course assignments.

The default policy of this course is that no late submissions will be accepted.

Note that I recognize exceptional circumstances may arise, and I am willing to work with students when they do. Therefore, the following additional requirements are put in place:

1. If you must travel for business and will have limited Internet connectivity, then you must notify the instructor at least one week prior to travel to make arrangements for handling assignments. Failure to provide this advanced notice will result in all due dates being enforced.
2. If you are traveling on vacation, then all due dates remain enforced. Personal travel is not an excuse to relax the due dates.
3. If there is a family emergency (e.g., a death in the family or a serious illness), then you must notify the instructor as soon as possible to make arrangements for handling the assignments. Note that this may require some form of documentation.
4. If you become personally ill, then it is important for you to take care of your health; however, since we are not meeting in person in a classroom, meaning that spread of disease is no longer an issue, only illnesses or injuries that require professional medical attention will receive special handling. Medical documentation may be requested; otherwise, all due dates will be enforced.
5. Under no circumstances will time management issues result in a relaxation of due dates. Poor time management is never an acceptable excuse.
6. Special accommodations are available for students who register a disability with the university.
   

Questions about this policy should be directed to the instructor.

Generative AI Policy

While research is often a collaborative process, the teaching of research is often very individually based. As such, it is essential to develop skills in the planning and conduct of research, as well as skills in the communication of, about, or around research. As such, the use of generative AI-based tools (e.g., Gemini or ChatGPT) is strictly prohibited in this course. That said, if you do use such tools, then you are required to cite that fact and furnish the transcript to the instructor. Failure to do so will be treated as plagiarism, which is a form of academic misconduct.

Academic Policies

Deadlines for Adding, Dropping and Withdrawing from Courses

Students may add a course up to one week after the start of the term for that particular course. Students may drop courses according to the drop deadlines outlined in the EP academic calendar (https://ep.jhu.edu/student-services/academic-calendar/). Between the 6th week of the class and prior to the final withdrawal deadline, a student may withdraw from a course with a W on their academic record. A record of the course will remain on the academic record with a W appearing in the grade column to indicate that the student registered and withdrew from the course.

Academic Misconduct Policy

All students are required to read, know, and comply with the Johns Hopkins University Krieger School of Arts and Sciences (KSAS) / Whiting School of Engineering (WSE) Procedures for Handling Allegations of Misconduct by Full-Time and Part-Time Graduate Students.

This policy prohibits academic misconduct, including but not limited to the following: cheating or facilitating cheating; plagiarism; reuse of assignments; unauthorized collaboration; alteration of graded assignments; and unfair competition. Course materials (old assignments, texts, or examinations, etc.) should not be shared unless authorized by the course instructor. Any questions related to this policy should be directed to EP’s academic integrity officer at ep-academic-integrity@jhu.edu.

Students with Disabilities - Accommodations and Accessibility

Johns Hopkins University values diversity and inclusion. We are committed to providing welcoming, equitable, and accessible educational experiences for all students. Students with disabilities (including those with psychological conditions, medical conditions and temporary disabilities) can request accommodations for this course by providing an Accommodation Letter issued by Student Disability Services (SDS). Please request accommodations for this course as early as possible to provide time for effective communication and arrangements.

For further information or to start the process of requesting accommodations, please contact Student Disability Services at Engineering for Professionals, ep-disability-svcs@jhu.edu.

Student Conduct Code

The fundamental purpose of the JHU regulation of student conduct is to promote and to protect the health, safety, welfare, property, and rights of all members of the University community as well as to promote the orderly operation of the University and to safeguard its property and facilities. As members of the University community, students accept certain responsibilities which support the educational mission and create an environment in which all students are afforded the same opportunity to succeed academically. 

For a full description of the code please visit the following website: https://studentaffairs.jhu.edu/policies-guidelines/student-code/

Classroom Climate

JHU is committed to creating a classroom environment that values the diversity of experiences and perspectives that all students bring. Everyone has the right to be treated with dignity and respect. Fostering an inclusive climate is important. Research and experience show that students who interact with peers who are different from themselves learn new things and experience tangible educational outcomes. At no time in this learning process should someone be singled out or treated unequally on the basis of any seen or unseen part of their identity. 
 
If you have concerns in this course about harassment, discrimination, or any unequal treatment, or if you seek accommodations or resources, please reach out to the course instructor directly. Reporting will never impact your course grade. You may also share concerns with your program chair, the Assistant Dean for Diversity and Inclusion, or the Office of Institutional Equity. In handling reports, people will protect your privacy as much as possible, but faculty and staff are required to officially report information for some cases (e.g. sexual harassment).

Course Auditing

When a student enrolls in an EP course with “audit” status, the student must reach an understanding with the instructor as to what is required to earn the “audit.” If the student does not meet those expectations, the instructor must notify the EP Registration Team [EP-Registration@exchange.johnshopkins.edu] in order for the student to be retroactively dropped or withdrawn from the course (depending on when the "audit" was requested and in accordance with EP registration deadlines). All lecture content will remain accessible to auditing students, but access to all other course material is left to the discretion of the instructor.