Research

Current Projects

• Advisors: Prof. Victoria Booth and Prof. Denise Kirschner.

• Developing a gradient-based parameter inference pipeline in PyTorch to calibrate a stochastic ODE model of polyphasic rodent sleep-wake dynamics against empirical EEG/EMG recordings.

• Designed a domain-informed summary statistic loss function, resolving fundamental optimization disconnects present in Maximum Mean Discrepancy (MMD)-based approaches for high-dimensional stochastic time series. 

• Automated hyperparameter optimization using Optuna's Tree-structured Parzen Estimator (TPE), reducing search time from weeks to ~20 hours of compute. 

• Applied iterative density-based calibration (CaliPro) and Neural Spline Flow simulation-based inference (SBI) to characterize posterior distributions over model parameters and reduce parameter search space by 88%. 

Publications

• Axinn, E. et al. "Orthogonal Polynomials on Bubble-Diamond Fractals." Complex Analysis and Operator Theory, 2025. https://link.springer.com/article/10.1007/s11785-025-01782-8.

Past Projects

• Probability of Tree Changes in the Ancestral Recombination Graph

Project for the Indiana University REU (Summer 2023). Derived closed-form expressions for transition probabilities of genealogical tree topologies under the ancestral recombination graph using martingale and Markov chain methods.

• WNBA Player Valuation & Salary Cap Optimization

Project for the Erdós Institute Data Science Boot Camp (Awarded With Distinction, Summer 2026). Built a Random Forest regression model to predict WNBA player fair market value (FMV) salaries from on-court performance statistics.