Blogs
Research stories behind some of my selected publications.
# Evolutionary Computation
# Surrogate Modeling
# Optimization
2025
Many real-world optimization problems come with a heavy price tag per evaluation — think expensive simulations or costly physical experiments. This post breaks down our three-stage surrogate-assisted differential evolution, which learns cheap models of the fitness landscape to find high-quality solutions while spending far fewer expensive evaluations.
# Large-Scale Optimization
# Decomposition
# Differential Evolution
2024
When a problem has thousands of variables, the search space is overwhelming. Differential grouping splits large problems into smaller, near-independent subproblems so an optimizer can conquer them one at a time — and this work makes the decomposition step itself dramatically faster.
# Deep Learning
# Clustering
# Representation Learning
2026
Contrastive clustering learns clusters directly from data without labels, but two silent failures — class collision and representation collapse — can quietly ruin the learned features. This work targets both with pseudo-supervised triplets and InfoNCE-style losses.
# Computer Vision
# Image Segmentation
# Object Tracking
2022
Mean-shift finds the modes of a density for segmentation and tracking, but it can be slow. GridShift makes mode-seeking dramatically faster by restricting the search to a coarse grid — without giving up accuracy.
# Power Systems
# Microgrids
# Load Flow
2021
Islanded microgrids lack the main grid’s voltage reference, so conventional power-flow tools fail. This post explains our inversion-free, robust load-flow algorithm that keeps microgrid power-flow analysis reliable.
# Metaheuristics
# Global Optimization
2019
Most evolutionary algorithms perturb variables in Cartesian space. Spherical search instead samples within a sphere — a surprisingly effective way to explore and refine solutions, and the seed of a whole family of algorithms.