Kiran Vodrahalli

{about} {research} {talks} {blog} {archive} {teaching} {notes} {links} {press}

research

Also see my Google Scholar and ArXiv pages (these may be incomplete).

Some of my papers have authors listed in alphabetical order (standard for theory papers).
Therefore, I use * to denote equal authorship to avoid confusion.

Preprints

[P5] Michelangelo: Long Context Evaluations Beyond Haystacks via Latent Structure Queries.
Kiran Vodrahalli, Santiago Ontanon, Nilesh Tripuraneni, Kelvin Xu, Sanil Jain, Rakesh Shivanna,
Jeffrey Hui, Nishanth Dikkala, Mehran Kazemi, Bahare Fatemi, Rohan Anil, Ethan Dyer,
Siamak Shakeri, Roopali Vij, Harsh Mehta, Vinay Ramasesh, Quoc Le, Ed Chi, Yifeng Lu, Orhan Firat,
Angeliki Lazaridou, Jean-Baptiste Lespiau, Nithya Attaluri, Kate Olszewska.
[arXiv] [Google DeepMind]

[P4] Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context.
Gemini Team, Google. Core Contributor.
[arXiv] [Google DeepMind]

[P3] Gemini: A Family of Highly Capable Multimodal Models.
Gemini Team, Google. Core Contributor.
[arXiv] [Google DeepMind Tech Report] [Google Blog] [Gemini Webpage]

[P2] PaLM 2 Technical Report.
Google. Core Contributor to Long Context workstream.
[arXiv] [Google AI Tech Report] [Google Blog] [PaLM 2 Webpage]

[P1] Nonlinear Initialization Methods for Low-Rank Neural Networks.
Kiran Vodrahalli, Rakesh Shivanna, Maheswaran Sathiamoorthy, Sagar Jain, Ed H. Chi.
[arXiv]

Conference and Journal Publications

All Publications

[C11] Online Learning with Bounded Recall.
Jon Schneider*, Kiran Vodrahalli*.
International Conference on Machine Learning, July 2024. Poster.
[arXiv] [code] [conference]

[C10] Is Learning in Games Good for the Learners?.
William Brown, Jon Schneider, Kiran Vodrahalli.
Neural Information Processing Systems, December 2023. Spotlight.
[pdf] [arXiv]

[C9] The Platform Design Problem.
Christos Papadimitriou*, Kiran Vodrahalli*, Mihalis Yannakakis*.
WINE Conference on Internet and Web Economics, December 2021. Oral Presentation and Poster.
[pdf] [arXiv] [poster] [conference]

[J2] Learning and Planning with Logical Automata.
Brandon Araki, Kiran Vodrahalli, Thomas Leech, Cristian-Ioan Vasile, Mark Donahue, Daniela Rus.
Autonomous Robots, August 2021.
[pdf] [journal]

[C8] The Logical Options Framework.
Brandon Araki, Xiao Li, Kiran Vodrahalli, Jonathan DeCastro, J. Micah Fry, Daniela Rus.
ICML International Conference on Machine Learning, July 2021. Long Oral Presentation and Poster.
[pdf] [mlr press] [poster] [icml]

[C7] Deep Bayesian Nonparametric Learning of Rules and Plans from Demonstrations with a Learned Automaton Prior.
Brandon Araki, Kiran Vodrahalli, Thomas Leech, Cristian-Ioan Vasile, Mark Donahue, Daniela Rus.
AAAI Conference on Artificial Intelligence, February 2020. Spotlight Presentation and Poster.
[pdf] [supplement] [conference] [poster]

[C6] Privacy Accounting and Quality Control in the Sage Differentially Private ML Platform.
Mathias Lécuyer, Riley Spahn, Kiran Vodrahalli, Roxana Geambasu, Daniel Hsu.
Symposium on Operation Systems Principles, October 2019. Oral Presentation.
[pdf] [arXiv] [conference] [poster]

[C5] Learning to Plan with Logical Automata.
Brandon Araki*, Kiran Vodrahalli*, Thomas Leech, Cristian-Ioan Vasile, Mark Donahue, Daniela Rus.
Robotics: Science and Systems, June 2019. Spotlight Presentation and Poster.
[pdf] [conference] [poster]

[C4] Attribute-Efficient Learning of Monomials over Highly-Correlated Variables.
Alex Andoni*, Rishabh Dudeja*, Daniel Hsu*, Kiran Vodrahalli*.
Algorithmic Learning Theory, March 2019. Oral Presentation.
[pdf] [pmlr] [conference] [poster]

[C3] A Large Self-Annotated Corpus for Sarcasm.
Mikhail Khodak, Nikunj Saunshi, Kiran Vodrahalli.
Language Resources and Evaluation, May 2018. Poster.
[pdf] [conference] [arXiv] [dataset] [code]

[C2] A Compressed Sensing View of Unsupervised Text Embeddings, Bag-of-n-Grams, and LSTMs.
Sanjeev Arora*, Mikhail Khodak*, Nikunj Saunshi*, Kiran Vodrahalli*.
International Conference on Learning Representations, April 2018. Poster.
[pdf] [blog] [abstract] [conference] [poster] [embedding code] [recovery code] [word vectors]

[C1] A Temporal Decay Model for Mapping between fMRI and Natural Language Annotations.
Kiran Vodrahalli, Cathy Chen, Viola Mocz, Christopher Baldassano, Uri Hasson, Sanjeev Arora, Kenneth A. Norman.
Cognitive Computational Neuroscience, September 2017. Poster.
[pdf] [conference] [poster] [code]

[J1] Mapping between fMRI responses to movies and their natural language annotations.
Kiran Vodrahalli, Po-Hsuan Chen, Yingyu Liang, Christopher Baldassano, Janice Chen,
Christopher Honey, Uri Hasson, Peter Ramadge, Kenneth A. Norman, Sanjeev Arora.
Neuroimage, June 2017.
[pdf] [journal] [arXiv] [code]

Workshop Publications

[W7] The Platform Design Problem.
Christos Papadimitriou*, Kiran Vodrahalli*, Mihalis Yannakakis*.
Strategic ML 2021 NeurIPS Workshop, December 2021. Spotlight Oral Presentation (top 10%) and Poster.
(Note: same work as WINE ‘21 conference publication).
[pdf] [arXiv] [poster]

[W6] The Platform Design Problem.
Christos Papadimitriou*, Kiran Vodrahalli*, Mihalis Yannakakis*.
EC 2021 NetEcon Workshop, July 2021. Oral Presentation and Poster.
(Note: same work as WINE ‘21 conference publication).
[pdf] [arXiv] [poster]

[W5] Learning to Plan with Logical Automata.
Brandon Araki*, Kiran Vodrahalli*, Cristian-Ioan Vasile, Daniela Rus.
NeurIPS 2018 Workshop on Infer2Control, December 2018. Oral Presentation and Poster.
(Note: same work as RSS ‘19 conference publication).
[pdf] [slides] [poster]

[W4] A Compressed Sensing View of Unsupervised Text Embeddings, Bag-of-n-Grams, and LSTMs.
Sanjeev Arora*, Mikhail Khodak*, Nikunj Saunshi*, Kiran Vodrahalli*.
ACL Workshop on Representation Learning for NLP, July 2018. Poster.
(Note: same work as ICLR ‘18 conference publication).

[W3] A Compressed Sensing View of Unsupervised Text Embeddings, Bag-of-n-Grams, and LSTMs.
Sanjeev Arora*, Mikhail Khodak*, Nikunj Saunshi*, Kiran Vodrahalli*.
ICML Workshop on Theory of Deep Learning, July 2018. Oral Presentation and Poster.
(Note: same work as ICLR ‘18 conference publication).

[W2] Mapping between Natural Movie fMRI Responses and Word-Sequence Representations.
Kiran Vodrahalli, Po-Hsuan Chen, Yingyu Liang, Janice Chen, Esther Yong,
Christopher Honey, Peter J. Ramadge, Kenneth A. Norman, Sanjeev Arora.
NeurIPS Workshop on Representation Learning in Artificial and Biological Neural Networks,
Dec 2016. Oral Presentation and Poster.
(Note: Earlier version of NeuroImage ‘17 journal publication).
[slides] [poster] [code]

[W1] A Semantic Shared Response Model.
Kiran Vodrahalli, Po-Hsuan Chen, Janice Chen, Esther Yong, Christopher Honey,
Peter J. Ramadge, Kenneth A. Norman, Sanjeev Arora.
ICML Workshop on Multi-view Representation Learning , Jun 2016. Oral Presentation and Poster.
(Note: Earlier version of NeuroImage ‘17 journal publication).
[pdf] [slides] [poster] [code]

Technical Reports and Theses

[T12] Resource-Efficient Methods in Machine Learning. Ph.D. Thesis (June 2022). Advised by Alex Andoni and Daniel Hsu. [pdf] [library]

[T11] Temporally Dependent Mappings Between fMRI Responses and Natural Language Descriptions of Natural Stimuli. COS MSE Thesis (May 2017). Advised by Sanjeev Arora and Ken Norman. [pdf] [code]

[T10] Low-dimensional Representations of Semantic Context in Language and the Brain. MAT AB Thesis (May 2016). Advised by Sanjeev Arora and Ken Norman. [pdf] [code]

[T9] Learning the Optimal Step Size for Gradient Descent on Convex Quadratics. Poster for NYAS ML Symposium 2020.
Alex Andoni*, Daniel Hsu*, Tim Roughgarden*, Kiran Vodrahalli*. [poster]

[T8] Can Simple Assembly Algorithms Compute Robust, High-Dimensional Means?. COMS 6998-06 Project (Fall 2018). Advised by Christos Papadimitriou. [pdf]

[T7] An Efficient General Algorithm for Interactive Clustering. COMS 6998-04 Project (Fall 2017). Advised by Daniel Hsu. [pdf]

[T6] Sparse, Low-dimensional and Multimodal Representations of Time Series for Mind-Reading. COS 513 Project (Fall 2015). With Lydia Liu and Niranjani Prasad. Advised by Barbara Engelhardt. [pdf] [blog]

[T5] Learning Shifting Communities Online in the Adversarial Block Model. APC 529 Project (Fall 2015). Advised by Emmanuel Abbe. [pdf]

[T4] Solving Word Analogies With Convex Optimization. COS 511 Project (Spring 2015). Advised by Elad Hazan. [pdf] [code]

[T3] Comparing Hebbian Semantic Vectors Across Language. NEU 330 Project (Spring 2015). Advised by Ken Norman. [pdf] [code]

[T2] Noun Compounds in Semantic Quad-Space. Junior Independent Work (Fall 2014). Advised by Christiane Fellbaum. [pdf] [code]

[T1] Estimating Trending Twitter Topics With Count-Min Sketch. COS 521 Project (Fall 2014). With Evan Miller and Albert Lee. Advised by Sanjeev Arora. [pdf] [code]

Code

For the code accompanying my research projects, see the links next to each paper or my github.

Coding Projects

Outershell. [code] Outershell is a simple hack to enable the bash shell with the full functionality of an iPython shell. The idea was inspired by chat windows with a computational environment: e.g., it would be awesome if one could interactively code with a chat buddy, with the data structures and functions remaining around for indefinite use. Outershell implements this kind of functionality for a single user on a computer. This kind of functionality already kind of exists in the form of the shell or various programs like Emacs and Vim (for power users at least), but Outershell democratizes its accessibility by enabling users to code their functions in Python.