2024
Mechanism for feature learning in neural networks and backpropagation-free machine learning models [Link]
Adityanarayanan Radhakrishnan*, Daniel Beaglehole*, Parthe Pandit, Mikhail Belkin.
Science.
Quadratic Models for Understanding Neural Network Dynamics [Link]
Libin Zhu, Chaoyue Liu, Adityanarayanan Radhakrishnan, Mikhail Belkin
ICLR.
2023
Transfer learning with kernel methods [Link]
Adityanarayanan Radhakrishnan*, Max Ruiz Luyten*, Neha Prasad, Caroline Uhler.
Nature Communications 14, Article 5570.
Wide and deep networks achieve consistency in classification [Link]
Adityanarayanan Radhakrishnan, Mikhail Belkin, Caroline Uhler.
PNAS 120, Article 14.
A cross-modal autoencoder framework learns holisitic representations of cardiovascular state [Link]
Adityanarayanan Radhakrishnan*, Samuel Freesun Friedman*, Shaan Khurshid, Kenney Ng, Puneet Batra, Steven Lubitz, Anthony Philippakis, Caroline Uhler.
Nature Communications 14, Article 2436.
2022
Simple, fast, and flexible framework for matrix completion with infinite width neural networks [Link]
Adityanarayanan Radhakrishnan, Mikhail Belkin, Caroline Uhler
PNAS 119, Article 16.
2021
Causal network models of SARS-CoV-2 expression and aging to identify candidates for drug repurposing [Link]
Anastasiya Belyaeva*, Louis Cammarata*, Adityanarayanan Radhakrishnan*, Chandler Squires, Karren Dai Yang, G.V. Shivashankar, Caroline Uhler
Nature Communications 12, Article 1024.
Local Quadratic Convergence of Stochastic Gradient Descent with Adaptive Step Size [arXiv]
Adityanarayanan Radhakrishnan, Mikhail Belkin, Caroline Uhler
ICML Workshop on Beyond first-order methods in ML systems.
Do Deeper Convolutional Networks Perform Better? [arXiv]
Eshaan Nichani*, Adityanarayanan Radhakrishnan*, Caroline Uhler.
ICML Workshop on Over-parameterization: Pitfalls and Opportunities.
On Alignment in Deep Linear Neural Networks [arXiv]
Adityanarayanan Radhakrishnan*, Eshaan Nichani*, Daniel Bernstein, Caroline Uhler
ICML Workshop on Over-parameterization: Pitfalls and Opportunities.
Multi-domain translation between single-cell imaging and sequencing data using autoencoders [Link]
Karren Yang, Anastasiya Belyaeva, Saradha Venkatachalapathy, Karthik Damodaran, Abigail Katcoff, Adityanarayanan Radhakrishnan, G.V. Shivashankar, Caroline Uhler
Nature Communications 12, Article 31.
2020
Overparameterized autoencoders implement associative memory [Link]
Adityanarayanan Radhakrishnan, Mikhail Belkin, Caroline Uhler
PNAS 119, Article16.
2019
Memorization in Overparameterized Autoencoders [arXiv, openreview]
Adityanarayanan Radhakrishnan, Mikhail Belkin, Caroline Uhler
ICML Workshop on Identifying and Understanding Deep Learning Phenomena.
2018
Patchnet: Interpretable Neural Net- works for Image Classification [arXiv]
Adityanarayanan Radhakrishnan, Charles Durham, Alican Soylemezoglu, Caroline Uhler
NeurIPS Machine Learning for Health (ML4H) Workshop.
Counting Markov equivalence classes for DAG models on trees [arXiv, DOI]
Adityanarayanan Radhakrishnan, Liam Solus, Caroline Uhler
Discrete Applied Mathematics.
2017
Machine learning for nuclear mechano-morphometric biomarkers in cancer diagnosis [Link]
Adityanarayanan Radhakrishnan, Karthik Damodaran, Alican Soylemezoglu, Caroline Uhler, G.V. Shivashankar
Scientific Reports 7, Article 17946.
Counting Markov equivalence classes by number of immoralities [arXiv, workshop link]
Adityanarayanan Radhakrishnan, Liam Solus, Caroline Uhler
UAI Special Workshop on Causality.
Preprints
Synthetic lethality screening with Recursive Feature Machines [bioRxiv]
Cathy Cai*, Adityanarayanan Radhakrishnan*, Caroline Uhler.
Linear Recursive Feature Machines provably recover low-rank matrices [Link]
Adityanarayanan Radhakrishnan, Dmitriy Drusvyatskiy, Mikhail Belkin.
Mechanism of feature learning in convolutional neural networks [arXiv]
Daniel Beaglehole*, Adityanarayanan Radhakrishnan*, Parthe Pandit, Mikhail Belkin.
A Mechanism for Producing Aligned Latent Spaces with Autoencoders [arXiv]
Saachi Jain*, Adityanarayanan Radhakrishnan*, Caroline Uhler.
Linear Convergence of Generalized Mirror Descent with Time-dependent Mirrors [arXiv]
Adityanarayanan Radhakrishnan, Mikhail Belkin, Caroline Uhler.
Adit Radha 