This page contains info and FAQs for prospective PhD students of Chris Donahue’s yet-to-be-named lab in the Computer Science Department at CMU.
My lab will generally work at the intersection of ML, music, multimedia, and creativity (notable venues include ISMIR, ICLR, ICML, NeurIPS, ICASSP, and CHI). At a high-level, my research involves (1) improving ML methods for generative modeling of music and other multimedia, and (2) building real-world interactive systems which allow a broad audience to harness the creative potential of generative models. A concrete example is my Piano Genie system which uses generative models to allow non-musicians a glimpse of musical improvisation. Please see my recent research statement* for a more thorough overview.
*Since I wrote this research statement one year ago, I have been fascinated by the ongoing “renaissance” in generative modeling for creativity—notably in the text-to-image space through models like DALL-E 2 and Stable Diffusion, but surely for many other forms of multimedia in the near future. I would also be excited to foster a lab that can be part of this conversation, both from a technical and societal perspective. On the technical side, I have particular interest in enabling real-time interactions for this class of large generative models.
At CMU, the School of Computer Science (SCS) has many departments, one of which is (confusingly) named the Computer Science Department (CSD). I am affiliated w/ CSD, so make sure to apply to CSD via the “Ph.D. in Computer Science” option to ensure your application will be routed properly:
You should also specify that you want to work with me directly. Because I don’t formally start until next fall, I am currently not listed on the list for specifying faculty of interest, though I am working on getting added (please check back). For now, please specify “Machine Learning” / “Other” for this section, and mention my name verbatim in your statement of purpose.
I am most directly looking for students w/ a technical background in ML, computer science, or engineering, as well as a background in some creative domain of interest and a desire to explore the use of ML in that domain. Ideally, you may already have some research experience or even publications. But these are very loose requirements and I am easily intrigued by passionate people from interesting backgrounds—please apply if you have any interest regardless of your background! I especially encourage students from groups historically underrepresented in computer science to apply.
Note that CSD admits students to the department rather than to individual labs. Hence, I can weigh in on admissions, but the final decisions are up to the department.
Hopefully pretty cool! Though many details may not materialize until next fall. Generally, I will strive to foster a diverse, inclusive, and fun group. I plan to lean into the interdisciplinary nature of our group and have regular lab meetings open not only to my students but also to interested folks from other groups (e.g., from Roger Dannenberg’s group) and departments (e.g., MLD, HCII, LTI, music). I am hoping to co-advise many of my students with other amazing faculty here at CMU. I also hope to be able to retain an affiliation w/ Google Magenta, which may allow my students to work on research projects at Google for a duration.
Yes, you may contact me via the email address on my personal homepage, though I may not be able to respond to all inquiries. Yonatan Bisk has a page w/ general wisdom about emailing faculty and the graduate admissions process.
If you have a general question about the application process, you can also open an issue on this GitHub page, which will have a higher likelihood of a response.