PhD Program in Machine Learning
The Ph.D. Program in Machine Learning is for students who are interested in research in Machine Learning and Computational Statistics. The program is operated jointly by faculty in the School of Computer Science and Department of Statistics.
The extraordinary spread of computers and online data is changing forever the way that important decisions are made in many organizations. Hospitals now analyze online medical records to decide which treatments to apply to future patients, banks analyze past financial records to learn to spot future fraud, and factories analyze past operations to learn to produce higher quality goods. Scientific research in many fields is also undergoing significant change as a result of dramatic increases in online data.
Understanding the most effective ways of using the vast amounts of data that are now being stored is a significant challenge to society, and therefore to science and technology, as it seeks to obtain a return on the huge investment that is being made in computerization and data collection. Advances in the development of automated techniques for data analysis and decision making requires interdisciplinary work in areas such as machine learning algorithms, the statistical and computational principles that underly these algorithms, database and data warehousing methods, complexity analysis, data visualization, privacy and security issues, and application areas such as business, marketing, and public policy.
Carnegie Mellon University's doctoral program in Machine Learning is designed to train students to become tomorrow's leaders in this rapidly growing area. The program is part of CMU's Machine Learning Department which is made up of a multi-disciplinary team of faculty and students across several academic departments. Machine Learning is dedicated to furthering scientific understanding of automated learning and to producing the next generation of tools for data analysis and decision making based on that understanding.
Today's demand for expertise in machine learning far exceeds the supply, and this imbalance will become more severe over the coming decade. Through a combination of interdisciplinary coursework, hands-on applications, and cutting-edge research, graduates of the Ph.D. program in Machine Learning will be uniquely positioned to pioneer new developments in this field, and to be leaders in both industry and academia.
Overview of Ph.D. Program Requirements
- Completion of required courses, Data Analysis Project, Conference Presentation Skills, and MS degree within 3 years (although many students do it in 2 or 2.5 years).
- Mastery of proficiencies in Teaching, Conference Presentation, and Research skills.
- Successful defense of a Ph.D. thesis.
The curriculum for the Machine Learning Ph.D. is built on a foundation of five core courses and one elective (plus the Data Analysis Project requirement). These six courses also comprise the required courses for the MS degree. Together with the Data Analysis Project requirement, these should be completed during the first three years of study.
A typical full-time, graduate course load during the first two years consists each term of two classes (at 12 graduate units per class) plus 24 units of advanced research. Thus, during the first two years, a student has the opportunity to take several elective classes in addition to the six required courses.
The ML curriculum joins courses with a Computer Science main theme and those with a Probability and Statistics main theme. These may be grouped, as follows:
In CS, relevant sub-fields include: Databases, Machine Learning, Data Mining, and Algorithms applications in areas such as Robotics, Information Retrieval, and AI.
In Statistics (including Philosophy), the sub-fields include: Statistical Modeling (e.g., hierarchical and times series), Bayes' Nets, Causation, and Experimental Design. The curriculum is based on core academic courses on Intermediate Statistics, Machine Learning, Statistical Machine Learning & Discovery, Multimedia Databases, and Algorithms.
These 3 required core courses together provide a foundation in machine learning, statistics, probability, and algorithms:
- 10-715 Advanced Introduction to Machine Learning
- 10-702 Statistical Machine Learning
- 36-705 Intermediate Statistics
Plus any 2 of the following menu of core courses:
- 10-708 Probabilistic Graphical Models
- 10-725 Convex Optimization
- 10-807 Deep Learning or 10-703 Deep Reinforcement Learning
- 15-750 Algorithms or 15-853 Algorithms in the Real World
- 15-850 Graduate Artificial Intelligence
- 15-826 Multimedia Databases and Data Mining
- 36-752 Advanced Probability
Plus 1 elective:
- An additional course from the Menu Core list above
- Any course at the 700 or higher level in SCS or Statistics (36-xxx)
- Other courses by approval
Note: Some students will have taken some of the above courses before entering the MLD PhD program: for example, as MS students at CMU. If students have previously taken the above-named courses at Carnegie Mellon before joining the MLD PhD, those may be used to satisfy the requirements and do not need to be repeated. (Note that courses can only be used for a single Master's degree.)
Some students will have taken similar courses at other universities before entering the MLD PhD program. Based on such equivalent coursework, any student can apply to replace (not reduce) up to two courses with either menu cores or electives. All requests must be supported by the advisor, and will be evaluated by the PhD co-directors.
Here is a typical schedule for the first two years of study:
FALL - 1st Year
SPRING - 1st Year
|10-715 Adv. Machine Learning||10-702 Statistical Machine Learning|
|36-705 Intermediate Statistics||Core course or Elective|
|10-920 Research (24 units)||10-920 Research (24 units)|
FALL - 2nd Year
SPRING - 2nd Year
|Core course or Elective||Core course or Elective|
|10-821 DAP Prep (6 units)||10-910 PhD DAP Research (12 units)|
|10-920 Research (30 units)||10-920 Research (24 units)|
The Third Year
By the third year, a Ph.D. student should have completed all coursework and the DAP. For those students seeking an academic position after completing the ML PhD, or those pursuing certain subfields, additional advanced electives in the allied disciplines of SCS, MCS, Philosophy, the Tepper School of Business, or others may be chosen in consultation with the student's advisor. As in each of the first two years, any coursework is supplemented by research, for a total of 48 units/term.
The Fourth Year and Beyond
A Ph.D. student typically presents a thesis proposal no later than the start of the fourth year, and then spends the fourth and sometimes fifth year working on their thesis research.
Data Analysis Project
During the second year, a Ph.D. student is required to demonstrate data analysis and machine learning skills in the context of a focused project. The Data Analysis Project may be carried out either at Carnegie Mellon or at a sponsoring corporate institution under the joint supervision of the sponsor and a ML faculty. It will be concluded by a written report (in lieu of a Master's Thesis) in which the student demonstrates an ability to approach data mining problems in a way that cuts across existing disciplinary boundaries. The requirement includes poster presentation and 5-minute oral presentation at DAP Day and also the submission of a DAP Paper. Passing this requirement will be the judgment of the DAP committee.
Student must form an official "DAP committee" to evaluate the document. The committee will consist of the advisor, another faculty member from any department, and an optional third member. The optional third member is typically someone with an interest in the analysis of the data set, and does not have to be an expert in ML, part of the student's thesis committee, or be faculty. The student should form the committee as early as possible during the DAP research process, with the committee completely formed no later than the beginning of the semester in which the DAP will be presented, and inform Diane Stidle of who the members are. The full committee is required to attend the presentation.
Conference Presentation Skills
During their second or third year, Ph.D. students must give a talk at least 30 minutes long, and invite members of the Presentation Skills committee to attend and evaluate it. As preparation for this requirement, students are invited to attend optional workshops on presentation skills, which will be offered twice in each semester (or more often, by demand).
Ph.D. students are required to serve as Teaching Assistants for two semesters in Machine Learning courses (10-xxx), beginning in their second year. This fulfills their Teaching Skills requirement.
It is expected that all Ph.D. students engage in active research from their first semester. Moreover, advisor selection occurs in the first month of entering the Ph.D. program, with the option to change at a later time. Roughly half of a student's time should be allocated to research and lab work, and half to courses until these are completed.
In addition, students must follow all university policies and procedures.
Machine Learning is committed to providing full tuition and stipend support for the academic year, for each full-time ML Ph.D. student, for a period of 5 years. Research opportunities are constrained by funding availability. ML's funding commitments assume that the student is making satisfactory progress in the program, as reported to the student at the end of each academic term. Students are strongly encouraged to compete for outside fellowships and other sources of financial support. ML will supplement these outside awards in order to fulfill its obligations for tuition and stipend support.
GRE General test scores are required for the application. You must speak English well: if you are not a native speaker, we recommend a combined TOEFL score of 100, with no subscore below 25, although we will make exceptions to this cutoff in exceptional cases. Unofficially, we recommend a high level of comfort with math (particularly linear algebra, probability, and proofs) and computer programming (at the level of an undergraduate degree in computer science, although many of our applicants get the necessary experience without majoring in CS). It is possible to fill in some of this background on the fly, but you will be working hard to do so! In addition, the program is very competitive, so successful applications always stand out in some way from their peers -- for example grades, research experience, or recommendation letters.