CAP 4611 - Algorithms for machine learning

Fall 2023

Course description:	An overview of the most commonly used algorithms for supervised, unsupervised, and reinforcement learning. Introduction to experimental design, evaluation metrics, and applications of machine learning.
Instructor:	Dr. Lotzi Bölöni
Office Location:	HEC - 319
E-mail:	Ladislau.Boloni@ucf.edu (preferred means of communication)
TAs:	TA: Ilkin Sevgi Isler ilkinsevgi.isler@ucf.edu Graders: Deepak Kumar Kunda de831452@ucf.edu Vamsi Krishna Yamani va364022@ucf.edu Aditya Kumar Aditya.Kumar@ucf.edu
Web Site:	http://www.cs.ucf.edu/~lboloni/Teaching/CAP4611_Fall2023/index.html The assignments and the other announcements will be posted on the course web site
Classroom:	CB2-0207
Class hours:	Mon, Wed 1:30PM - 2:45PM
Office hours:	Mon, Wed 5:00pm - 6:30pm (in HEC 319)
Enrollment requirements:	COP 3502C and STA 2023 each with a grade of C (2.0) or better. While not required, knowledge of multivariable calculus, linear algebra, and probability are also strongly recommended.
Required texts:	There is no required textbook.
Grading methods:	Grading scale: Letter grades only (A,B,C and F) The standard 90 / 80 / 70 scale will be applied. Homeworks 1-6: 10% each, Midterm 1,2 and Final 10% each, Class participation / microquizzes: 10%. Homeworks will be open 1 week past deadline, with a 20% penalty. No further extensions will be given. The exams will be administered through the Respondus Lockdown Browser, and are open book, open notes. Make up exams will be given only in justified cases.
Academic integrity	Students should familiarize themselves with UCF's Rules of Conduct at . According to Section 1, “Academic Misconduct,” students are prohibited from engaging in 1. Unauthorized assistance: Using or attempting to use unauthorized materials, information or study aids in any academic exercise unless specifically authorized by the instructor of record. The unauthorized possession of examination or course-related material also constitutes cheating. 2. Communication to another through written, visual, electronic, or oral means: The presentation of material which has not been studied or learned, but rather was obtained through someone else's efforts and used as part of an examination, course assignment, or project. 3. Commercial Use of Academic Material: Selling of course material to another person, student, and/or uploading course material to a third-party vendor without authorization or without the express written permission of the university and the instructor. Course materials include but are not limited to class notes, Instructor's PowerPoints, course syllabi, tests, quizzes, labs, instruction sheets, homework, study guides, handouts, etc. 4. Falsifying or misrepresenting the student's own academic work. 5. Plagiarism: Using or appropriating another's work without any indication of the source, thereby attempting to convey the impression that such work is the student' s own. 6. Multiple Submissions: Submitting the same academic work for credit more than once without the express written permission of the instructor. 7. Helping another violate academic behavior standards. 8. Soliciting assistance with academic coursework and/or degree requirements. Responses to Academic Dishonesty, Plagiarism, or Cheating Students should also familiarize themselves with the procedures for academic misconduct in UCF's student handbook, The Golden Rule . UCF faculty members have a responsibility for students' education and the value of a UCF degree, and so seek to prevent unethical behavior and respond to academic misconduct when necessary. Penalties for violating rules, policies, and instructions within this course can range from a zero on the exercise to an “F” letter grade in the course. In addition, an Academic Misconduct report could be filed with the Office of Student Conduct, which could lead to disciplinary warning, disciplinary probation, or deferred suspension or separation from the University through suspension, dismissal, or expulsion with the addition of a “Z” designation on one's transcript. Being found in violation of academic conduct standards could result in a student having to disclose such behavior on a graduate school application, being removed from a leadership position within a student organization, the recipient of scholarships, participation in University activities such as study abroad, internships, etc. Let's avoid all of this by demonstrating values of honesty, trust, and integrity. No grade is worth compromising your integrity and moving your moral compass. Stay true to doing the right thing: take the zero, not a shortcut. Unauthorized Use of Websites and Internet Resources There are many websites claiming to offer study aids to students, but in using such websites, students could find themselves in violation of academic conduct guidelines. These websites include (but are not limited to) Quizlet, Course Hero, Chegg Study, and Clutch Prep. UCF does not endorse the use of these products in an unethical manner, which could lead to a violation of our University's Rules of Conduct. They encourage students to upload course materials, such as test questions, individual assignments, and examples of graded material. Such materials are the intellectual property of instructors, the university, or publishers and may not be distributed without prior authorization. Students who engage in such activity could be found in violation of academic conduct standards and could face course and/or University penalties. Please let me know if you are uncertain about the use of a website so I can determine its legitimacy. Unauthorized Distribution of Class Notes Third parties may attempt to connect with you to sell your notes and other course information from this class. Distributing course materials to a third party without my authorization is a violation of our University's Rules of Conduct. Please be aware that such class materials that may have already been given to such third parties may contain errors, which could affect your performance or grade. Recommendations for success in this course include coming to class on a routine basis, visiting me during my office hours, connecting with the Teaching Assistant (TA), and making use of the Student Academic Resource Center (SARC), the University Writing Center (UWC), the Math Lab, etc. If a third party should contact you regarding such an offer, I would appreciate your bringing this to my attention. We all play a part in creating a course climate of integrity.
In-class recording	Students may, without prior notice, record video or audio of a class lecture for a class in which the student is enrolled for their own personal educational use. A class lecture is defined as a formal or methodical oral presentation as part of a university course intended to present information or teach enrolled students about a particular subject. Recording class activities other than class lectures, including but not limited to lab sessions, student presentations (whether individually or part of a group), class discussion (except when incidental to and incorporated within a class lecture), clinical presentations such as patient history, academic exercises involving student participation, test or examination administrations, field trips, private conversations between students in the class or between a student and the faculty member, and invited guest speakers is prohibited. Recordings may not be used as a substitute for class participation and class attendance, and may not be published or shared without the written consent of the faculty member. Failure to adhere to these requirements may constitute a violation of the University's Student Code of Conduct as described in the Golden Rule.
Course accessibility:	The University of Central Florida is committed to providing access and inclusion for all persons with disabilities. Students with disabilities who need access to course content due to course design limitations should contact the professor as soon as possible. Students should also connect with Student Accessibility Services (Ferrell Commons 185, sas@ucf.edu, phone 407-823-2371). For students connected with SAS, a Course Accessibility Letter may be created and sent to professors, which informs faculty of potential course access and accommodations that might be necessary and reasonable. Determining reasonable access and accommodations requires consideration of the course design, course learning objectives and the individual academic and course barriers experienced by the student. Further conversation with SAS, faculty and the student may be warranted to ensure an accessible course experience.
Campus safety statement:	Emergencies on campus are rare, but if one should arise during class, everyone needs to work together. Students should be aware of their surroundings and familiar with some basic safety and security concepts. In case of an emergency, dial 911 for assistance. Every UCF classroom contains an emergency procedure guide posted on a wall near the door. Students should make a note of the guide's physical location and review the online version at https://centralflorida-prod.modolabs.net/student/safety/index. Students should know the evacuation routes from each of their classrooms and have a plan for finding safety in case of an emergency. If there is a medical emergency during class, students may need to access a first-aid kit or AED (Automated External Defibrillator). To learn where those are located, see https://ehs.ucf.edu/automated-external-defibrillator-aed-locations. To stay informed about emergency situations, students can sign up to receive UCF text alerts by going to https://my.ucf.edu and logging in. Click on “Student Self Service” located on the left side of the screen in the toolbar, scroll down to the blue “Personal Information” heading on the Student Center screen, click on “UCF Alert”, fill out the information, including e-mail address, cell phone number, and cell phone provider, click “Apply” to save the changes, and then click “OK.” Students with special needs related to emergency situations should speak with their instructors outside of class. To learn about how to manage an active-shooter situation on campus or elsewhere, consider viewing this video https://youtu.be/NIKYajEx4pk.
Deployed active duty military students	If you are a deployed active duty military student and feel that you may need a special accommodation due to that unique status, please contact your instructor to discuss your circumstances.

Syllabus

Date	Topic	Lecture Notes, Readings, Homeworks
Mon, Aug. 21	Introduction to machine learning What is machine learning? Relationship to artificial intelligence Relationship to statistics Supervised learning Unsupervised learning Reinforcement learning	[slides] Introduction to Machine Learning [homework] HW1: Predictions about machine learning - due August 25, 2023
Wed, Aug. 23	Review of the math for ML vectors and matrices	[slides] Math review
Mon, Aug. 28	Review of the math for ML (cont'd) partial derivatives & gradients	[homework] HW2: Programming tools of machine learning: jupyter, numpy, pandas, matplotlib - ~~Due Sept 11~~ Due Sept 18 [homework] Dataset for HW2
Wed, Aug. 30	Hurricane Idalia
Mon, Sep. 4	Labor day - no class
Wed, Sept. 6	Programming tools of machine learning numpy: vector and matrix operations pandas: manipulating datasets matplotlib: plotting data	[slides] Programming tools
Mon, Sep. 11	Aquiring, loading and cleaning the data Data types Where to get public datasets Preprocessing and cleaning the data Preliminary exploration of the data What do we want to predict? Choosing features Noise	[slides] Data
Wed, Sep. 13
Mon, Sep. 18	Linear regression Linear regression in single variable Cost function Gradient descent	[slides] Linear regression - single variable
Tue, Sep. 20	Linear regression - multivariable Gradient descent in multiple features Feature engineering Normal equation	[slides] Linear regression - multiple features [homework] HW3: Linear regression - Due October 4, 2023
Mon, Sept 25	Logistic regression Classification Sigmoid cost, probability based output Multi-class classification	[slides] Logistic regression
Wed, Sep. 27	Midterm 1: Data cleaning + Linear regression
Mon, Oct 2	Overfitting and regularization Overfitting view Regularization The hypothesis selection view. Training data, validation data, test data. Trends when underfitting Trends when not training enough Trends when overfitting Regularization	[slides] Overfitting and regularization
Wed, Oct. 4	Overfitting and regularization cont'd
Mon, Oct. 9	K-nearest neighbors K-nearest neighbors algorithm Minkowski distance Problem of dimensionality Embedding space	[slides] k nearest neighbors
Wed, Oct. 11	Decision trees Relationship to expert systems Impurity functions ID3 algorithm CART classification and regression trees	[slides] Decision trees
Mon, Oct. 16	Bias and variance Understanding expectations Understanding sampling The bias/variance tradeoff	[slides] The bias / variance tradeoff
Wed, Oct. 18	Bagging Ensemble models Bootstrap aggregating Random forests	[slides] Bagging
Mon, Oct. 23	Boosting Weak learners Boosting AdaBoost	[slides] Boosting
Wed, Oct. 25	Unsupervised learning overview Self-supervision techniques Finding structure in unsupervised data	[slides] Unsupervised intro. PCA. [homework] HW4: Nearest neighbors - Due Nov 8 [homework] HW5: Ensemble methods - Due Nov 15
Mon, Oct. 30	Midterm 2: up to Boosting (inclusive)
Wed, Nov. 1	Clustering K-means clustering Applications lecture by Ilkin Sevgi Isler	[slides] K-means clustering
Mon, Nov. 6	Introduction to deep learning and neural networks History, perceptrons, linearity Nonlinearity: backpropagation General formulation: loss functions	[slides] Introduction to neural networks and deep learning
Wed, Nov. 8	Fully connected networks	[slides] Fully connected networks
Mon, Nov. 13	Convolutional neural networks Properties of the visual domain Convolutions Convolutional layers Pooling CNN architectures for image classification	[slides] Convolutional neural networks [slides] Convolutions as an image processing technique [slides] Convolutional layers
Wed, Nov. 15		[homework] HW6: Neural networks - Due Dec. 2
Mon, Nov 20	Recurrent neural networks Feedforward vs recurrent The problem of memory Long stort term memory (LSTM) Applications in robotics	[slides] Recurrent neural networks [slides] Applications in robotics
Wed, Nov 22	Thanksgiving break - no class
Mon, Nov. 27	Attention and transformers The problem of attention in series The transformer architecture Applications: large language models	[slides] Transformers
Wed, Nov. 29	Artificial General Intelligence The Turing test The alignment problem Societal impacts of human-scale AGI Superintelligence	[slides] Artificial General Intelligence - Pros and Cons
Dec. 4	Final exam: Monday December 4, 2023 1:00-3:50pm