Resume

I enjoy creating new things, improving existing methods, inventing new approaches, and solving challenging problems.

I have a PhD in Mathematics and am a Machine Learning Research Scientist at Lirio. Currently I'm interested in the mathematical aspects of machine learning, especially in relation to reinforcement learning and artificial intelligence. Below is a breakdown of my professional experiences and accomplishments.

For more information, check out a detailed discussion of my research interests, a list of non-professional side projects, or download a short version of my resume.

resume.pdf

Work Experience

Machine Learning Research Scientist
2020.03 – Present

Lirio LLC
Behavioral Reinforcement Learning Lab
Lirio AI Research
Knoxville, TN, USA

My main responsibility at Lirio is performing independent research on topics that are beneficial to the operation of the Lirio Behavioral Nudging Agent. Our particular topics of interest include:

Reinforcement learning for hyperpersonalization
Policy evaluation and ranking under limited data
Nonconvex gradient-free optimization
Behavioral modeling for human decision-making

During my time at Lirio (so far), I have published 3 papers and given 4 talks at conferences and workshops. Due to the impact of our research, Lirio won the Best AI Startup Award by The AIconics in 2020 and received the Excellence in AI Award by The Business Intelligence Group in 2022. I was promoted to Senior Machine Learning Research Scientist in 2021.

Postdoctoral Research Associate
2017.09 – 2020.02

Oak Ridge National Laboratory
Computing and Computational Sciences
Computer Science and Mathematics Division
Oak Ridge, TN, USA

As a Postdoctoral Research Associate at ORNL, I conducted research that supported the Department of Energy initiatives. The particular topics of my research included:

Mathematical aspects of supervised learning
Neural architecture design and initialization
Reduced order modeling in non-Hilbert spaces

During my time at ORNL, I published 5 papers and gave 4 talks at conferences and workshops.

Graduate Research Assistant
2016.08 – 2017.08

University of South Carolina
College of Arts and Sciences
Department of Mathematics
Columbia, SC, USA

As a Graduate Research Assistant at USC, I conducted research that was relevant to the topic of my dissertation. My particular topics of interest included:

High-dimensional constructive approximation
Greedy algorithms with computational relaxation
Sparse convex optimization in Banach spaces

During my time at USC, I published 2 papers and presented at 3 conferences and workshops. I was awarded the Dean's Dissertation Fellowship in April of 2016.

Education

PhD in Mathematics
2012.08 – 2017.08

University of South Carolina
College of Arts and Sciences
Department of Mathematics
Columbia, SC, USA

diploma.pdf

dissertation.pdf

BA in Mathematics
2007.09 – 2012.06

Lomonosov Moscow State University
Faculty of Mechanics and Mathematics
Department of Mathematical Analysis
Moscow, Russia

msu_diploma.pdf

thesis.pdf

Publications

Below is a list of completed papers that are either published or under review. I also have a bunch of half-finished papers for most of my ongoing research directions and some for my side projects.

Data-Centric Approach to Constrained Machine Learning: A Case Study on Conway’s Game of Life
A. Bibin, A. Dereventsov
submitted to 38th Annual AAAI Conference on Artificial Intelligence
arXiv (will be uploaded by 12/23)
Gaussian Smoothing Gradient Descent for Minimizing High-Dimensional Non-Convex Functions
A. Starnes, A. Dereventsov, C. Webster
submitted to SIAM Journal on Optimization
arXiv
Zero-Shot Recommendations with Pre-Trained Large Language Models for Multimodal Nudging
R. Harrison, A. Dereventsov, A. Bibin
accepted to International Conference on Data Mining Workshops (ICDMW23), IEEE, 2023
arXiv
Increasing Entropy to Boost Policy Gradient Performance on Personalization Tasks
A. Starnes, A. Dereventsov, C. Webster
accepted to International Conference on Data Mining Workshops (ICDMW23), IEEE, 2023
arXiv
Examining Policy Entropy of Reinforcement Learning Agents for Personalization Tasks
A. Dereventsov, A. Starnes, C. Webster
submitted to ACM Transactions on Recommender Systems
arXiv
Offline Policy Comparison under Limited Historical Agent-Environment Interactions
A. Dereventsov, J. Daws, C. Webster
accepted to RAMSES workshop at SISSA
arXiv
The Natural Greedy Algorithm for Reduced Bases in Banach Spaces
A. Dereventsov, C. Webster
submitted to Numerische Mathematik
arXiv
Simulated Contextual Bandits for Personalization Tasks from Recommendation Datasets
A. Dereventsov, A. Bibin
International Conference on Data Mining Workshops (ICDMW22), IEEE, 2022
Paper, arXiv
Modeling Non-deterministic Human Behaviors in Discrete Food Choices
A. Starnes, A. Dereventsov, S. Blazek, F. Phillips
International Conference on Data Mining Workshops (ICDMW22), IEEE, 2022
Paper, arXiv
Biorthogonal Greedy Algorithms in Convex Optimization
A. Dereventsov, V. Temlyakov
Applied and Computational Harmonic Analysis, 60, 2022
Paper, arXiv
On the Unreasonable Efficiency of State Space Clustering in Personalization Tasks
A. Dereventsov, R. Vatsavai, C. Webster
International Conference on Data Mining Workshops (ICDMW21), IEEE, 2021
Paper, arXiv
An Adaptive Stochastic Gradient-Free Approach for High-Dimensional Blackbox Optimization
A. Dereventsov, C. Webster, J. Daws
International Conference on Computational Intelligence (ICCI20), Springer, 2022
Paper, arXiv
Greedy Shallow Networks: An Approach for Constructing and Training Neural Networks
A. Dereventsov, A. Petrosyan, C. Webster
International Journal of Artificial Intelligence, 19(2), 2021
Paper, arXiv
Neural Network Integral Representations with the ReLU Activation Function
A. Petrosyan, A. Dereventsov, C. Webster
Mathematical and Scientific Machine Learning Conference, PMLR, 107, 2020
Paper, arXiv
A Unified Way of Analyzing Some Greedy Algorithms
A. Dereventsov, V. Temlyakov
Journal of Functional Analysis, 277(12), 2019
Paper, arXiv
On the Generalized Approximate Weak Chebyshev Greedy Algorithm
A. Dereventsov
Studia Mathematica, 237(2), 2017
Paper, arXiv
On the Approximate Weak Chebyshev Greedy Algorithm in Uniformly Smooth Banach Spaces
A. Dereventsov
Journal of Mathematical Analysis and Applications, 436(1), 2016
Paper, arXiv
Comparison of the Convergence Rate of Pure Greedy and Orthogonal Greedy Algorithms
A. Dereventsov
Mathematical Notes, 92(3), 2012
Paper

Conferences and Workshops

Below is a list of conferences and workshops where I have presented my research (at least the ones I can remember). I still have the slides for each talk, and in some cases even the recording. Please reach out if you're interested in any of them!

2022.11 ICDM Workshop on AI for Nudging and Personalization (WAIN22), Orlando, FL, USA
Talk 1: Simulated Contextual Bandits for Personalization Tasks from Recommendation Datasets
Talk 2: Modeling Non-deterministic Human Behaviors in Discrete Food Choices
2021.12 ICDM Workshop on AI for Nudging and Personalization (WAIN21), virtual
Talk: On the Unreasonable Efficiency of State Space Clustering in Personalization Tasks
2021.09 “Approximation and Discretization”, Moscow, Russia
Talk: High-Dimensional Approximation via the Smoothing-Based Gradient-Free Optimization
2021.03 SIAM Conference on Computational Science and Engineering (CSE21), virtual
Talk: Adaptive Stochastic Gradient Free Optimization for Reinforcement Learning
2020.12 International Conference on Computational Intelligence (ICCI20), virtual
Talk: An Adaptive Stochastic Gradient-Free Approach for High-Dimensional Blackbox Optimization
2019.09 43rd SIAM SEAS Annual Meeting, Knoxville, TN, USA
Talk 1: Greedy Shallow Networks: A New Approach for Constructing Neural Networks
Talk 2: Natural Greedy Algorithm: Constructing Reduced Bases in Banach Spaces
2019.07 AI Expo at Oak Ridge National Laboratory, Oak Ridge, TN, USA
Poster: Reduced Basis Learning via the Natural Greedy Algorithm
2019.05 16th International Approximation Theory Conference, Nashville, TN, USA
Talk: The Natural Greedy Algorithm for Reduced Bases in Banach Spaces
2017.06 “Approximation Theory and Function Spaces” workshop at CRM, Barcelona, Spain
Talk: The Chebyshev Greedy Algorithm with Imprecise Step Evaluations
2017.05 International Conference in Approximation Theory, Savannah, GA, USA
Poster: The Chebyshev Greedy Algorithm with Imprecise Step Evaluations
2016.05 15th International Approximation Theory Conference, San Antonio, TX, USA
Talk: On the Generalized Approximate Weak Chebyshev Greedy Algorithm
2012.01 “Contemporary Problems of the Function Theory and its Applications”, Saratov, Russia
Talk: Comparison of Convergence Rates of the Pure Greedy and Orthogonal Greedy Decompositions
2011.01 “Modern Methods in Theory of Functions and Adjacent Problems”, Voronezh, Russia
Talk: On the Convergence Rates of the Orthogonal Greedy Decompositions

Resume

Work Experience

Machine Learning Research Scientist2020.03 – Present

Postdoctoral Research Associate2017.09 – 2020.02

Graduate Research Assistant2016.08 – 2017.08