Ph.D. Student Toyota Technological Institute at Chicago Email: [firstname] [at] ttic [dot] edu Address: 6045 S Kenwood, Room 512-13, Chicago, IL-60637.

About

I am a second-year Ph.D. student at the Toyota Technological Institute at Chicago (TTIC), advised by Nathan Srebro.

My research is on computational learning theory and related topics: complexity and algorithms for learning, probability theory, statistics, and optimization. More generally, I have borad interests in artificial intelligence, theoretical computer science, and information theory.

Preprints/Working Projects

• Learning under Group Invaraince: SQ Lower Bounds via Harmonic Analysis

• Learning Iterated Sequence-to-Next-Token Predictors with Chain-of-Thoughts

• Exact Community Recovery under Side Information: Optimality of Spectral Algorithms

Publications

• On the Complexity of Learning Sparse Functions with Statistical and Gradient Queries
  NeurIPS 2024
  Nirmit Joshi, Theodor Misiakiewicz, Nathan Srebro

• Noisy Interpolation Learning with Shallow Univariate ReLU Networks
  ICLR 2024 (Spotlight)
  Nirmit Joshi, Gal Vardi, Nathan Srebro

• The Limits and Potentials of Local SGD for Distributed Heterogeneous Learning with Intermittent Communication
  COLT 2024
  Kumar Kshitij Patel, Margalit Glasgow, Ali Zindari, Lingxiao Wang, Sebastian U. Stich, Ziheng Cheng, Nirmit Joshi, Nathan Srebro

• Community Detection in the Hypergraph SBM: Exact Recovery Given the Similarity Matrix
  COLT 2023
  Julia Gaudio, Nirmit Joshi

• Generalizing Greenwald-Khanna Streaming Quantile Summaries for Weighted Inputs
  ICDT 2023
  Sepehr Assadi, Nirmit Joshi, Milind Prabhu, Vihan Shah