|
|
General Information | Resources | Weekly Schedule | Lecture Notes |
I. General Information |
Dr. Yoonsuck Choe
Email: choe(a)tamu.edu
Office: HRBB 322B
Phone: 845-5466
Zoom (office hour): http://tamu.zoom.us/my/yoonsuckchoe
Zoom (lectures): see Canvas https://canvas.tamu.edu/courses/45731
Office hours: Tue/Thu 3pm-4pm (Zoom: see above)
Qing Wan (TA)
Email: frankwan(a)email.tamu.edu
Office hours: Wed/Fri 3-4pm.
Zoom: https://tamu.zoom.us/my/frankwan Ankit Garg (Grader)
Email: ankitgarg(a)tamu.edu
* Graders do not hold office hours.
Grade of C or better in MATH 304 and STAT 211; grade of C or better in CSCE 221 or STAT 404.
TR 8:00am-9:15am, ONLINE, Synchronous (you will receive the zoom link via email)
Machine learning is the study of self-modifying computer systems that can acquire new knowledge and improve their own performance; survey machine learning techniques, which include induction from examples, Bayesian learning, artificial neural networks, instance-based learning, genetic algorithms, reinforcement learning, unsupervised learning, and biologically motivated learning algorithms. Prerequisite: CSCE 420 or CSCE 625.
The goal of this course is to help you to
- learn the theoretical foundations of machine learning,
- learn various problems and solution strategies in machine learning, and
- learn practical methodology for applying ML algorithms to problem domain of your choice.
The expected accomplishments of the students are as follows:
- Become confident in the basic mathematics and algorithms underlying machine learning.
- Become confident in applying machine learning techniques to solve novel problems.
- Become effective in rapidly prototyping machine learning algorithms.
- Become effective in designing and conducting experiments with machine learning algorithm, and analyzing the results.
The students who take this course should be able to demonstrate the following upon the completion of this course.
- Knowledge of basic mathematics and algorithms underlying various machine learning techniques.
- Knowlege of various supervised, unsupervised, and reinforcement learning algorithms.
- Ability to implement, from scratch, the above mentioned learning algorithms.
- Ability to formulate problems in a machine learning framework, and select the most appropriate machine learning algorithm.
- Ability to design and conduct machine learning experiments.
- Ability to analyze the results from machine learning experiments.
- Appreciation of latest advances in machine learning (including but not limited to biologically motivated, neuroscience-inspired machine learning methods).
- Ability to seek more advanced knowledge in machine learning, as needed.
- Ability to use Colab for rapid prototyping, experimenting, and documenting.
- Main text (required): Ethem Alpaydin (2014) Introduction to Machine Learning, 3rd edition, MIT Press. [Book home page (3rd edition)] [Book home page (2nd edition)] [Book home page (1st edition)]
- Secondary text (optional but strongly recommended): Tom Mitchell (1997) Machine Learning, McGraw-Hill. [Book home page]
- Other recommended books:
- Kevin P. Murphy, "Probabilistic Machine Learning: An Introduction", MIT press, 2021. [Book web site, with downloadable full text]
- Chris Bishop, "Pattern Recognition and Machine Learning", Springer, 2006. [Book web site].
- Ian Goodfellow, Yoshua Bengio, and Aaron Courville, "Deep Learing", MIT press, 2016. [Book web site]
- Aston Zhang, Zachary C. Lipton, Mu Li, and Alesander J. Smola, "Dive into Deep Learning", [Book web site, with downloadable jupyter notebook files]
See the Weekly Schedule section for more details. The content will closely reflect a combination of Alpaydin + Mitchell.
- computational learning theory: sample complexity, version space, PAC learning, VC dimension, mistake bound, cross-validation
- neural networks: perceptrons, multilayer perceptrons, backpropagation, gradient descent, stochastic gradient descent, optimizers, deep neural networks (convolutional networks, recurrent networks, attention)
- decision tree learning, support vector machines,
- dimensionality reduction: PCA, factor analysis, multidimensional scaling, Isomap, locally linear embedding
- genetic algorithms, neuroevolution
- unsupervised learning and local models: competitive learning, k-means, adaptive resonance theory, self-organizing maps, radial basis functions, regression, learning vector quantization.
- Bayesian learning: maximum likelihood, minimum description length, Bayes optimal classifiers, Naive bayes classifiers. Bayesian belief network, EM algorithm.
- Biologically motivated models.
The cutoff for an `A' will be at most 90% of total score, 80% for a `B', 70% for a `C', and 60% for a `D'.
- 5 assignments (each includes written and programming components), 10% each = 50%.
- Late policy: 1 point (out of 100 point total) deduction per hour (24/100 per day).
- All submissions should be done as colab notebooks. Download the ipynb file and submit it to Canvas. After completing each problem in the homework, "pin" the revision. As a separate file, submit the revision logs. Hand-written notes may be included in the notebook (scan or photo).
- 2 exams, 20% each = 40%
- Exam will be conducted on Canvas.
- Attendance: 10%
- Attendance will be checked through zoom log.
- If you are absent from the class (either late or leave early) for more than 20 minutes, it will count as an absence.
- First absence will be excused (unconditionally, except for exams).
- Afterwards, each absence will result in 2% deduction.
- Note: absences will continue to deduct 2% from your total score, beyond the 10%, up to 30% (by this point, you've missed almost half of the semester).
- All work should be done individually and on your own.
- If you find solutions to homeworks or programming assignments on the web (or in a book, etc.), check with the instructor first to determine if you can use it or not.
- Assignments turned in that are significantly similar will be reported to the Aggie Honor System Office.
- There will be no make up exams unless it is due to a genuine emergency defined in university rules. Examples of events that do not count as an emergency include the following: (1) Merely visiting the doctor's office without an explicit note from the office requesting absence. (2) Interview trips that got scheduled at the last moment, etc. Note that this is not an exhaustive list. All make up exams, if given, will be different from the original exam. Instructor may choose to give an oral exam instead of a written exam.
III. Weekly Schedule and Class Notes |
|
|
|||||
1 | 1/19 | Introduction | Alpaydin chap 1; Mitchell 1.1–1.2, 1.3–1.5 | slide01-intro.pdf slide01b-intro.pdf web_link; |
||
1 | 1/21 | " | slide01-intro.pdf slide01b-intro.pdf web_link; |
|||
2 | 1/26 | Supervised Learning (general) | Alpaydin chap 2; Mitchell 7.1–7.2, 7.4 | slide02-sup.pdf |
||
2 | 1/28 | " | Homework 1 announced | slide02-sup.pdf |
||
3 | 2/2 | Multilayer perceptrons | Alpaydin chap 11; Mitchell chap 4 | slide03-nnet.pdf |
||
3 | 2/4 | " | slide03-nnet.pdf |
|||
4 | 2/9 | " | slide03-nnet.pdf |
|||
4 | 2/11 | " | slide03-nnet.pdf |
|||
5 | 2/16 | No class | Class canceled due to weather | |||
5 | 2/18 | No class | Class canceled due to weather (see hw1 tip videos and backprop video on Canvas) | |||
6 | 2/23 | Reinforcement learning | Alpaydin chap 18; Mitchell chap 13 | Homework 2 announced | slide04-rl.pdf |
|
6 | 2/25 | " | Homework 1 due 2/28 Sunday | slide04-rl.pdf |
||
7 | 3/2 | Texas Independence Day (No Class) | No class | |||
7 | 3/4 | Reinforcement Learning | Alpaydin chap 18; Mitchell chap 13 | slide04-rl.pdf |
||
8 | 3/9 | Midterm Exam (in class) | ||||
8 | 3/11 | Reinforcement Learning | Alpaydin chap 18; Mitchell chap 13 | slide04-rl.pdf |
||
9 | 3/16 | Decision tree learning | Alpaydin chap 9; Mitchell chap 3 | slide05-dt.pdf |
||
9 | 3/18 | Redefined day (no class) | Students attend friday classes. No class for TR classes. March 19: Spring break | Homework 3 announced | Homework 2 due Monday 3/22 | |
10 | 3/23 | Decision tree learning | Alpaydin chap 9; Mitchell chap 3 |
slide05-dt.pdf |
||
10 | 3/25 | Dimensionality reduction | Alpaydin chap 6: 6.1–3, 6.7, 6.8 | slide06-dim.pdf |
||
11 | 3/30 | Dimensionality reduction | Alpaydin chap 6: 6.1–3, 6.7, 6.8 | slide06-dim.pdf |
||
11 | 4/1 | Bayesian learning | Mitchell chap 6 | slide07-bayes.pdf |
||
12 | 4/6 | Bayesian learning | Mitchell chap 6 | Homework 4 announced | Homework 3 due (4/6 Tuesday) | slide07-bayes.pdf |
12 | 4/8 | Local models | Alpaydin chap 12 | slide08-local.pdf |
||
13 | 4/13 | Local models | Alpaydin chap 12 | slide08-local.pdf |
||
13 | 4/15 | Deep learning | slide09-dl.pdf slide09-dl-suppl.pdf |
|||
14 | 4/20 | Deep learning | slide09-dl.pdf slide09-dl-suppl.pdf |
|||
14 | 4/22 | Deep learning | slide09-dl.pdf slide09-dl-suppl.pdf |
|||
15 | 4/27 | Deep learning | slide09-dl.pdf slide09-dl-suppl.pdf |
|||
15 | 4/29 | Advanced topics | See Canvas-files-lecture slides; Support Vector Machine slide (informational purposes only) | Homework 4 due May 1 (Sat) | slide10-svm.pdf |
|
5/5 (Wed) | Final Exam (5:00pm-7:30pm) |
IV. University Policy |
The university views class attendance and participation as an individual student responsibility. Students are expected to attend class and to complete all assignments.Please refer to Student Rule 7 in its entirety for information about excused absences, including definitions, and related documentation and timelines.
Also, see local course policy regarding attendance.
Students will be excused from attending class on the day of a graded activity or when attendance contributes to a student's grade, for the reasons stated in Student Rule 7, or other reason deemed appropriate by the instructor.Please refer to Student Rule 7 in its entirety for information about makeup work, including definitions, and related documentation and timelines.
Absences related to Title IX of the Education Amendments of 1972 may necessitate a period of more than 30 days for make-up work, and the timeframe for make-up work should be agreed upon by the student and instructor (Student Rule 7, Section 7.4.1).
The instructor is under no obligation to provide an opportunity for the student to make up work missed because of an unexcused absence (Student Rule 7, Section 7.4.2).
Students who request an excused absence are expected to uphold the Aggie Honor Code and Student Conduct Code. (See Student Rule 24.)
AGGIE HONOR CODE: An Aggie does not lie, cheat, or steal or tolerate those who do.Texas A&M University students are responsible for authenticating all work submitted to an instructor. If asked, students must be able to produce proof that the item submitted is indeed the work of that student. Students must keep appropriate records at all times. The inability to authenticate one's work, should the instructor request it, may be sufficient grounds to initiate an academic misconduct case (Section 20.1.2.3, Student Rule 20).
You can learn more about the Aggie Honor System Office Rules and Procedures, academic integrity, and your rights and responsibilities at http://aggiehonor.tamu.edu.
If there are any questions or concerns about whether an action is appropriate, you should check with the professor or teaching assistant first. If in doubt, assume that it is not appropriate.
Texas A&M University is committed to providing equitable access to learning opportunities for all students. If you experience barriers to your education due to a disability or think you may have a disability, please contact Disability Resources in the Student Services Building or at (979) 845-1637 or visit disability.tamu.edu. Disabilities may include, but are not limited to attentional, learning, mental health, sensory, physical, or chronic health conditions. All students are encouraged to discuss their disability related needs with Disability Resources and their instructors as soon as possible.
Texas A&M University is committed to fostering a learning environment that is safe and productive for all. University policies and federal and state laws prohibit gender-based discrimination and sexual harassment, including sexual assault, sexual exploitation, domestic violence, dating violence, and stalking.With the exception of some medical and mental health providers, all university employees (including full and part-time faculty, staff, paid graduate assistants, student workers, etc.) are Mandatory Reporters and must report to the Title IX Office if the employee experiences, observes, or becomes aware of an incident that meets the following conditions (see University Rule 08.01.01.M1):
Mandatory Reporters must file a report regardless of how the information comes to their attention including but not limited to face-to-face conversations, a written class assignment or paper, class discussion, email, text, or social media post. Although Mandatory Reporters must file a report, in most instances, you will be able to control how the report is handled, including whether or not to pursue a formal investigation. The University's goal is to make sure you are aware of the range of options available to you and to ensure access to the resources you need.
- The incident is reasonably believed to be discrimination or harassment.
- The incident is alleged to have been committed by or against a person who, at the time of the incident, was (1) a student enrolled at the University or (2) an employee of the University.
Students wishing to discuss concerns in a confidential setting are encouraged to make an appointment with Counseling and Psychological Services (CAPS)
Students can learn more about filing a report, accessing supportive resources, and navigating the Title IX investigation and resolution process on the University's Title IX webpage.
Texas A&M University recognizes that mental health and wellness are critical factors that influence a student's academic success and overall wellbeing. Students are encouraged to engage in proper self-care by utilizing the resources and services available from Counseling and Psychological Services (CAPS). Students who need someone to talk to can call the TAMU Helpline (979-845-2700) from 4:00 p.m. to 8:00 a.m. weekdays and 24 hours on weekends. 24-hour emergency help is also available through the National Suicide Prevention Hotline (800-273-8255) or at suicidepreventionlifeline.org.
To promote public safety and protect students, faculty, and staff during the coronavirus pandemic, Texas A&M University has adopted policies and practices for the Fall 2020 academic term to limit virus transmission. Students must observe the following practices while participating in face-to-face courses and course-related activities (office hours, help sessions, transitioning to and between classes, study spaces, academic services, etc.):
- Self-monitoring: Students should follow CDC recommendations for self-monitoring. Students who have a fever or exhibit symptoms of COVID-19 should participate in class remotely and should not participate in face-to-face instruction.
- Face Coverings: Face coverings (cloth face covering, surgical mask, etc.) must be properly worn in all non-private spaces including classrooms, teaching laboratories, common spaces such as lobbies and hallways, public study spaces, libraries, academic resource and support offices, and outdoor spaces where 6 feet of physical distancing is difficult to reliably maintain. Description of face coverings and additional guidance are provided in the Face Covering policy and Frequently Asked Questions (FAQ) available on the Provost website.
- Physical Distancing: Physical distancing must be maintained between students, instructors, and others in course and course-related activities.
- Classroom Ingress/Egress: Students must follow marked pathways for entering and exiting classrooms and other teaching spaces. Leave classrooms promptly after course activities have concluded. Do not congregate in hallways and maintain 6-foot physical distancing when waiting to enter classrooms and other instructional spaces.
- To attend a face-to-face class, students must wear a face covering (or a face shield if they have an exemption letter). If a student refuses to wear a face covering, the instructor should ask the student to leave and join the class remotely. If the student does not leave the class, the faculty member should report that student to the Student Conduct office for sanctions. Additionally, the faculty member may choose to teach that day's class remotely for all students.
Students required to quarantine must participate in courses and course-related activities remotely and must not attend face-to-face course activities. Students should notify their instructors of the quarantine requirement. Students under quarantine are expected to participate in courses and complete graded work unless they have symptoms that are too severe to participate in course activities.Students experiencing personal injury or Illness that is too severe for the student to attend class qualify for an excused absence (See Student Rule 7, Section 7.2.2.) To receive an excused absence, students must comply with the documentation and notification guidelines outlined in Student Rule 7. While Student Rule 7, Section 7.3.2.1, indicates a medical confirmation note from the student's medical provider is preferred, for Fall 2020 only, students may use the Explanatory Statement for Absence from Class formin lieu of a medical confirmation. Students must submit the Explanatory Statement for Absence from Class within two business days after the last date of absence.
Last edited:2/20/21, 10:07AM (Sat)