Publications

Publications by Borna Khodabandeh in statistical machine learning, generative modelling, optimization, and adversarial robustness.

Publications

See also my Google Scholar profile. Authors marked with * contributed equally.

Accelerated Random-Sweep Gibbs Sampling for Gaussian Graphical Models via Dual Normal Factor Graphs

Borna Khodabandeh, Mehdi Molkaraie

Preprint (to be submitted to IEEE Transactions on Information Theory), 2026

arXiv link

Shows that Gibbs sampling convergence is significantly accelerated in the dual model of a Gaussian graphical model; derives exact convergence rates and an explicit algebraic relation between primal and dual covariance structures.

LORE: Lagrangian-Optimized Robust Embeddings for Visual Encoders

Borna Khodabandeh*, Amirabbas Afzali*, Amirhossein Afsharrad, Shahabeddin Mousavi, Sanjay Lall, Sajjad Amini, Seyed-Mohsen Moosavi-Dezfooli

NeurIPS 2025, 2025

arXiv link

Co-first author. Conducted as Bachelor’s project. Developed a Lagrangian optimization framework for adversarially robust visual encoders (CLIP, DinoV2), achieving state-of-the-art robustness while maintaining clean accuracy.

Optimizing Contrastive Learning Models via Preference Optimization

Amirabbas Afzali*, Borna Khodabandeh*, Ali Rasekh, Mahyar JafariNodeh, Sepehr Kazemi Ranjbar, Simon Gottschalk

ICLR 2025, 2024

arXiv link

Co-first author. Introduced policy optimization for fine-tuning contrastive models (e.g., CLIP) using human preferences; improves alignment and robustness against backdoor attacks and dataset biases.

Counter Histogram-Based Forensics using Mean Structural Similarity Index Metric

R. Kazemi, A. Amini, B. Khodabandeh, M. Alikhani

SIAM J. Imaging Sci., 2025, 2023

paper link