There are three categories of courses: Foundation (group A), Core (group B), and Elective and Research requirements (group C) for the graduate program. These course requirements are intended to ensure that students are exposed to (1) fundamental concepts and tools (Foundation), (2) advanced, up-to-date views in topics central to Data Science for all students (Core), and (3) a deep, current view of their research or application are (Elective). Courses may not fulfill more than one requirement.

The doctoral program is structured as a total of 52 units in courses from these group A, B, and C as described in detail here. Out of the 52 units, 48 units (or 12 courses) must be taken for letter grade and at least 40 units must be using graduate-level courses.

The remaining 4 (= 52 – 48) units are for professional preparation, consisting of 1 unit of faculty research seminar, 2 units of TA/tutor training and 1 unit of survival skills course taken for a passing (satisfactory) grade. Finally, as mentioned earlier, out of the 12 regular courses, at least 10 must be graduate-level courses; at most two can be upper-level undergraduate courses. 36 units or 9 courses must be completed within six quarters from the start of the degree program.

Research rotations provide the opportunity for first-year PhD students to obtain research experience in data analysis under the guidance of HDSI affiliated faculty members. Through the rotations, students can identify a faculty member under whose supervision their dissertation research will be completed.

A research rotation is a guided research experience lasting one quarter (10 weeks) obtained by registering for DSC 294 with an instructor. All Ph.D. students will participate in a minimum of 2 research rotations during their first year, and with a minimum of two different faculty members, and as much as three rotations including summer quarter. A student may rotate twice under the same faculty member as long as they rotate with at least two faculty members. The goal is to expose students to new methodological approaches or domain knowledge and help them find an advisor for their Ph.D. research.

Please refer to the HDSI Faculty page for a list of HDSI faculty members and their research interests. Students should reach out to faculty members to learn more about potential rotation opportunities.

An important goal of HDSI is to foster interdisciplinary collaboration. Specifically, collaboration between method researchers and domain researchers. Domain researchers specialize in a particular scientific field and collect or generate large datasets. Domain experts can partner with HDSI faculty and jointly advise a PhD. student. The site matchpoint.com was designed to facilitate the creation of such collaborations.

Outcome of each rotation program: A written report, or curated datasets with jupyter notebooks. Students are expected to work 10 hours per week.

Research rotations must be completed by the end of the Fall Quarter of the second year with a signed commitment form from a faculty advisor. Those who fail to identify a research advisor shall be advised to leave the doctoral program with an optional assessment for completion of a terminal MS-DS degree.

The goal of the preliminary assessment examination is to assess students’ preparation for pursuing a PhD in data science, in terms of core knowledge and readiness for conducting research. The preliminary assessment is an advisory examination.

There is an oral presentation that needs to be completed before the end of Fall quarter of the second year. The student will propose the topic of the presentation (e.g., the outcome of a research rotation or a literature survey). The Graduate Program Committee will set up a committee consisting of two members (students can suggest committee members but there is no guarantee that only those suggested faculty members will be chosen for the committee). The oral presentation from the student will be followed by a Q&A session by the committee members. The committee will assess both the oral presentation as well as the students academic performance so far (especially in the required core courses). Students who did not get a satisfactory evaluation will receive a recommendation from the Graduate Program Committee regarding ways to remedy the lacking preparation (e.g, suggestion of courses to be taken), or an opportunity to receive a terminal MS in Data Science degree provided the student can meet the degree requirements of the MS program.

A research qualifying examination (UQE) is conducted by the dissertation committee consisting of four or more members approved by the graduate division as per senate regulation 715(D). One senate faculty member must have a primary appointment in the department outside of HDSI. Faculty with 25% or less partial appointment in HDSI may be considered for meeting this requirement on an exceptional basis upon approval from the graduate division.

The goal of UQE is to assess the ability of the candidate to perform independent critical research as evidenced by a presentation and writing a technical report at the level of a peer-reviewed journal or conference publication. The examination is taken after the student and his or her adviser have identified a topic for the dissertation and an initial demonstration of feasible progress has been made. The candidate is expected to describe his or her accomplishments to date as well as future work. The research qualifying examination must be completed no later than fourth year or 12 quarters from the start of the degree program; the UQE is tantamount to the advancement to PhD candidacy exam.

A petition to the Graduate Committee is required for students who take UQE after the required 12 quarters deadline. Students who fail the research qualifying examination may file a petition to retake it; if the petition is approved, they will be allowed to retake it one (and only one) more time. Students who fail UQE may also petition to transition to a MS in Data Science track.

Students must successfully complete a final dissertation defense oral presentation and examination to the Dissertation Committee consisting of four or more members approved by the graduate division as per senate regulation 715(D). The primary Thesis Adviser, who will chair the Dissertation Committee, must be a senate faculty member with an appointment of 0% or more at HDSI. One senate faculty member in the Dissertation Committee must have a primary appointment in a department outside of HDSI. Partially appointed faculty in HDSI (at 25% or less) are acceptable in meeting this outside-department requirement as long as their main (lead) department is not HDSI.

A dissertation in the scope of Data Science is required of every candidate for the PhD degree. HDSI PhD program thesis requirements must meet Regulation 715(D) requirements. The final form of the dissertation document must comply with published guidelines by the Graduate Division.

The dissertation topic will be selected by the student, under the advice and guidance of Thesis Adviser and the Dissertation Committee. The dissertation must contain an original contribution of quality that would be acceptable for publication in the academic literature that either extends the theory or methodology of data science, or uses data science methods to solve a scientific problem in applied disciplines.

The entire dissertation committee will conduct a final oral examination, which will deal primarily with questions arising out of the relationship of the dissertation to the field of Data Science. The final examination will be conducted in two parts. The first part consists of a presentation by the candidate followed by a brief period of questions pertaining to the presentation; this part of the examination is open to the public. The second part of the examination will immediately follow the first part; this is a closed session between the student and the committee and will consist of a period of questioning by the committee members.

Special Requirements: Generalization, Reproducibility and Responsibility
A candidate for doctoral degree in data science is expected to demonstrate evidence of generalization skills as well as evidence of reproducibility in research results. Evidence of generalization skills may be in the form of — but not limited to — generalization of results arrived at across domains, or across applications within a domain, generalization of applicability of method(s) proposed, or generalization of thesis conclusions rooted in formal or mathematical proof or quantitative reasoning supported by robust statistical measures. Reproducibility requirement may be satisfied by additional supplementary material consisting of code and data repository. The dissertation will also be reviewed for responsible use of data.