anandpareshb[at]google[dot]com
anandbrahmbhatt27[at]gmail[dot]com
anand[dot]brahmbhatt[at]iitdalumni[dot]com
Research Interests
Learning Theory
Generalization
Differential Privacy
Fairness
Bio
I am a Pre-Doctoral Researcher at Google Research India, working in the Advertising Sciences Team mentored by Dr. Rishi Saket and Dr. Aravindan Raghuveer. My work focuses on developing an understanding of the learnability of aggregate data and the privacy afforded by forming such aggregations. My CV is available here.
I am broadly interested in Learning Theory, Generalization, Differential Privacy and Fairness in Machine Learning. I want to develop a more abstract understanding of learning algorithms and how these relate to deeper concepts in Mathematics.
Previously, I was a Research Intern at Adobe Research, Bengaluru. I worked with Dr. Shiv Kumar Saini and Dr. Atanu R Sinha on developing methods for fairer Cloud Resource Allocation.
I graduated from Indian Institute of Technology, Delhi in 2022. I completed my undergraduate thesis under the guidance of Prof. Parag Singla and Prof. Mausam. The focus of my research there was to identify reasons and figure out ways to mitigate “Bias Amplification in Deep Networks”. I was also mentored by Prof. Amitabha Tripathi with whom I did a summer project on Graph Labelings.
Updates
- [September 2023] Our paper PAC Learning Linear Thresholds from Label Proportions got accepted to NeurIPS 2023 as a Spotlight. I’ll be attending NeurIPS 2023 in New Orleans and will be presenting this work as a poster.
- [August 2023] US Patent approved for our work on Cloud Based Resource Allocation using Meters at Adobe Research.
- [July 2023] Attended Conference on Learning Theory (COLT), 2023 in Bengaluru.
- [July 2022] Joined Google Research India as Pre-Doctoral Researcher in the Ad Sciences Team.
- [May 2022] Completed my Bachelors in Computer Science and Engineering at IIT Delhi.
- [May 2021] Joined Adobe Research, Bengaluru as a Research Intern.
Publications & Patents
PAC Learning Linear Thresholds from Label Proportions
Spotlight Paper @ NeurIPS 2023
Paper
Label Differential Privacy via Aggregation
Preprint (under submission @ ITCS 2024)
Paper
LLP-Bench: A Large Scale Tabular Benchmark for Learning from Label Proportions
Preprint
Paper
Towards Fair and Calibrated Models
Preprint (under submission @ AAAI 2024)
Paper
Cloud Based Resource Allocation Using Meters
US Patent 20230259403
Patent
Measures for Closeness to Cordiality for Graphs
Preprint
Summary of Results