publications

2026

1. 

   Implicit Regularization of SGD Reduces Shortcut Learning

   Nahal Mirzaie, Alireza Alipanah, Ali Abbasi, Amirmahdi Farzane, Hossein Jafarinia, Erfan Sobhaei, Mahdi Ghaznavi, Amir Najafi, Mahdieh Soleymani Baghshah, and Mohammad Hossein Rohban

   In The Fourteenth International Conference on Learning Representations, 2026

   Abs Code

   Training with stochastic gradient descent (SGD) at moderately large learning rates has been observed to improve robustness against spurious correlations, strong cor- relation between non-predictive features and target labels. Yet, the mechanism underlying this effect remains unclear. In this work, we identify batch size as an additional critical factor and show that robustness gains arise from the implicit regularization of SGD, which intensifies with larger learning rates and smaller batch sizes. This implicit regularization reduces reliance on spurious or short- cut features, thereby enhancing robustness while preserving accuracy. Importantly, this effect appears unique to SGD: gradient descent (GD) does not con- fer the same benefit and may even exacerbate shortcut reliance. Theoretically, we establish this phenomenon in linear models by leveraging statistical formulations of spurious correlations, proving that SGD systematically suppresses spurious feature dependence. Empirically, we demonstrate that the effect extends to deep neural networks across multiple benchmarks.

2025

1. 

   Snuffy: Efficient Whole Slide Image Classifier

   Hossein Jafarinia, Alireza Alipanah, Saeed Razavi, Nahal Mirzaie, and Mohammad Hossein Rohban

   In Computer Vision – ECCV 2024, 2025

   Abs Code

   Whole Slide Image (WSI) classification with multiple instance learning (MIL) in digital pathology faces significant computational challenges. Current methods mostly rely on extensive self-supervised learning (SSL) for satisfactory performance, requiring long training periods and considerable computational resources. At the same time, no pre-training affects performance due to domain shifts from natural images to WSIs. We introduce Snuffy architecture, a novel MIL-pooling method based on sparse transformers that mitigates performance loss with limited pre-training and enables continual few-shot pre-training as a competitive option. Our sparsity pattern is tailored for pathology and is theoretically proven to be a universal approximator with the tightest probabilistic sharp bound on the number of layers for sparse transformers, to date. We demonstrate Snuffy’s effectiveness on CAMELYON16 and TCGA Lung cancer datasets, achieving superior WSI and patch-level accuracies. The code is available on https://github.com/jafarinia/snuffy.

2. 

   Navigating the MIL Trade-Off: Flexible Pooling for Whole Slide Image Classification

   Hossein Jafarinia, Danial Hamdi, Amirhossein Alamdar, Elahe Zahiri, Soroush Vafaie Tabar, Alireza Alipanah, Nahal Mirzaie, Saeed Razavi, Amir Najafi, and Mohammad Hossein Rohban

   In Advances in Neural Information Processing Systems, 2025

   Code

3. 

   The Silent Helper: How Implicit Regularization Enhances Group Robustness

   Nahal Mirzaie, Mahdi Ghaznavi, Hosna Oyarhoseini, Alireza Alipanah, Erfan Sobhaei, Ali Abbasi, Amirmahdi Farzane, Hossein Jafarinia, Parsa Sharifi Sedeh, Arefe Boushehrian, Mahdieh Soleymani Baghshah, and Mohammad Hossein Rohban

   In High-dimensional Learning Dynamics 2025, 2025

2024

1. 

   MILFORMER: Weighted Dual Stream Class Centered Random Attention Multiple Instance Learning for Whole Slide Image Classification

   Hossein Jafarinia, Danial Hamdi, Alireza Alipanah, Nahal Mirzaie, and Mohammad Hossein Rohban

   2024

   Abs DOI Code

   Whole Slide Imaging (WSI) in contemporary pathology is pivotal, particularly in cancer diagnosis and prognosis. Multiple Instance Learning (MIL) has emerged as a pivotal strategy to address the scarcity of localized annotations in WSI analysis. However, in the current landscape of state-of-the-art methods, the instance-level accuracy of these models significantly lags behind that of the bag-level. This article introduces MILFormer, a novel multi-head class-centered random global attention, enhancing the weighted averaged dual-stream architecture. Our method surpasses existing bag and WSI classification approaches while notably improving instance-level accuracy by up to 8%, marking a substantial advancement in Regions of Interest (ROI) detection. We comprehensively evaluated our method, spanning various datasets, including the public Camelyon16 dataset, and showed that our proposed architecture is effective and robust across a broad spectrum of MIL challenges. Our code is available at: https://github.com/rohban-lab/MILFormer

2. Workshop on Spurious Correlation and Shortcut Learning: Foundations and Solutions

   Hesam Asadollahzadeh, Mahdi Ghaznavi, Mahdieh Soleymani Baghshah, Mohammad Hossein Rohban, Shikai Qiu, Polina Kirichenko, Aahlad Manas Puli, Arash Mari Oriyad, Parsa Hosseini, and Nahal Mirzaie

   In ICLR 2025 Workshop Proposals, 2024

2023

1. 

   Weakly-Supervised Drug Efficiency Estimation with Confidence Score: Application to COVID-19 Drug Discovery

   Nahal Mirzaie, Mohammad V. Sanian, and Mohammad H. Rohban

   In Medical Image Computing and Computer Assisted Intervention – MICCAI 2023, 2023

   Abs Code

   The COVID-19 pandemic has prompted a surge in drug repurposing studies. However, many promising hits identified by modern neural networks failed in the preclinical research, which has raised concerns about the reliability of current drug discovery methods. Among studies that explore the therapeutic potential of drugs for COVID-19 treatment is RxRx19a. Its dataset was derived from High Throughput Screening (HTS) experiments conducted by the Recursion biotechnology company. Prior research on hit discovery using this dataset involved learning healthy and infected cells’ morphological features and utilizing this knowledge to estimate contaminated drugged cells’ scores. Nevertheless, models have never seen drugged cells during training, so these cells’ phenotypic features are out of their trained distribution. That being said, model estimations for treatment samples are not trusted in these methods and can lead to false positives. This work offers a first-in-field weakly-supervised drug efficiency estimation pipeline that utilizes the mixup methodology with a confidence score for its predictions. We applied our method to the RxRx19a dataset and showed that consensus between top hits predicted on different representation spaces increases using our confidence method. Further, we demonstrate that our pipeline is robust, stable, and sensitive to drug toxicity.

2019

1. Parameterized synthesis of self-stabilizing protocols in symmetric networks

   Nahal Mirzaie, Fathiyeh Faghih, Swen Jacobs, and Borzoo Bonakdarpour

   Dec 2019

   DOI
2. Parameterized Synthesis of Self-Stabilizing Protocols in Symmetric Rings

   Nahal Mirzaie, Fathiyeh Faghih, Swen Jacobs, and Borzoo Bonakdarpour

   In 22nd International Conference on Principles of Distributed Systems (OPODIS 2018), 2019

   DOI