Bijan Mazaheri

alt text 

Ph.D. Candidate, NSF Graduate Research Fellow
Department of Computing & Mathematical Sciences
California Institute of Technology

About me

I am a computer scientist and mathematician interested in hidden variables in causal structures: when and if they can be learned and how they affect networks of opinions. While machine learning is often focused on building models from data, my work explores what models can tell us about incomplete or absent data. I am currently pursuing a Ph.D. in Computing and Mathematical Sciences at Caltech supported by a NSF Graduate Research Fellowship and an Amazon AI4Science Fellowship.

I am also a competitive runner. My personal best for the marathon is 2:15:26 at the Chicago Marathon and I was one of the youngest athletes at the 2020 US Olympic Marathon Trials. In 2022, I was the USA 50k National Champion. For more information, see running info.


I work with Professor Jehoshua Bruck and Professor Leonard Schulman on topics related to machine learning including:
  • Expert Graphs: We are developing a structure to study consistency in overlapping expertise. This work helps us understand non-transitive properties in networks of machine learning classifiers as well as how to combine classifiers to synthesize new knowledge.
  • Source Identification: Controlled experiments and causal inference require an understanding of confounding variables. We use mixture models to learn unobservable confounding variables given assumptions about the visible variables. This allows us to perform causal inference without measuring, or even knowing about hidden confounders.
  • Covariate Shift: Preparing a machine learning classifier for demographic differences between training data and deployment. I am interested in developing stable and reliable methods for real world situations that diverge from the available training data.


  • S. Gordon, *B. Mazaheri, Y. Rabani, L. Schulman. "Identifying Discrete Mixtures of Bayesian Networks." [arXiv]
  • S. Jain, B. Mazaheri, N. Raviv, J. Bruck. “Glioblastoma signature in the DNA of blood-derived cells” PLoS ONE 16(9): e0256831. 2021. [PLoS ONE]
  • B. Mazaheri, S. Jain, J. Bruck. “Expert Graphs: Synthesizing New Expertise via Collaboration.” IEEE ISIT 2021. [arXiv]
  • S. Gordon, *B. Mazaheri, Y. Rabani, L. Schulman. "Source Identification for Mixtures of Product Distributions." COLT 2021. [arXiv]
  • B. Mazaheri, S. Jain, J. Bruck. “Robust Correction of Sampling Bias using Cumulative Distribution Functions.” NeurIPS 2020. [NeurIPS 2020, arXiv]
  • S. Gordon, *B. Mazaheri, Y. Rabani, L. Schulman. “The Sparse Hausdorff Moment Problem, with Applications to Topic Models.” 2020. [arXiv]
  • S. Jain, B. Mazaheri, N. Raviv, J. Bruck. “Cancer Classification from Healthy DNA using Machine Learning.” 2019. [bioRxiv, Patent]
  • S. Jain, B. Mazaheri, N. Raviv, J. Bruck. “Short Tandem Repeats Information in TCGA is Statistically Biased by Amplification.” 2019. [bioRxiv]
* = Co-authorship


  • Cross Country Rankings - I maintain a website that uses statistical modeling to adjust cross country times based on the difficulty of the course. The website is used to generate rankings and simulate hypothetical races. Since its release in September, the website has had over 400,000 views and 10,000 unique users. The data has been used for a number of smaller research projects by students across the US.


  • California Institute of Technology - Ph.D., Computing and Mathematical Sciences
  • Cambridge University (Emmanuel College) - Studied mathematics under a Herchel Smith Fellowship
  • Williams College - B.A., physics and computer science