Research software and supporting materials

Julia and GPU

cuHALLaR

GPU-accelerated low-rank augmented Lagrangian solver for large-scale semidefinite programming.

The GPU variant of HALLaR is geared toward structured, large SDP instances and is currently the clearest public code artifact tied to my recent work.

Julia and CPU

HALLaR

Low-rank augmented Lagrangian solver for large-scale semidefinite programming.

HALLaR is the CPU-based predecessor to cuHALLaR and remains a useful reference point for the underlying hybrid low-rank approach.

Benchmarks and data

Transportation Networks

Transportation research datasets used for traffic assignment experiments.

This repository is the public support material most relevant to the bicriterion traffic assignment work listed on the research page.

Large-scale LPs

Hybrid Approaches for Large-Scale Linear Programs

Implementation work for hybrid methods targeting large linear programs with exploitable structure.

Traffic assignment

BiCriteria Traffic Assignment

Solver and experiment code for bicriterion traffic assignment and associated computational studies.

First-order methods

Accelerated Smoothing Methods

Experimental code accompanying work on accelerated smoothing and first-order complexity guarantees.