Deep learning researcher · PhD student
MohammadDehghanmanshadi
I develop learning systems at the intersection of machine learning, computer vision, and medical image analysis.
Currently pursuing my PhD at Iowa State University, with an interest in robust and responsible AI for real-world scientific and clinical problems.
Education
- B.S. · Electrical EngineeringBabol Noshirvani University of Technology (NIT)
- M.S. · Electrical EngineeringIran University of Science and Technology (Elm o Sanat)
- M.S. · Computer ScienceIowa State University (Cyclone)
Research focus
Reliable learning from complex visual data.
01
Medical image analysis
02
AI reliability & testing
03
Representation learning
04
Domain adaptation
05
Learning with limited labels
Recent
News
Our diffusion-based domain adaptation work for cell counting was published at IEEE ICMLA.
Defended my M.S. thesis in Computer Science at Iowa State University.
Our team shared the top rank in a real-world cell-counting competition supported by the National Science Foundation.
Started my PhD at Iowa State University.
Ranked 5th among more than 1,200 applicants in Iran’s national AI PhD entrance examination.
Selected work
Publications
- 01
InST-Microscopy method: style-token learning and stylized image generation. - 02
Taxonomy of MRI segmentation methods for missing modalities. Medical Image Segmentation on MRI Images with Missing Modalities: A Review
- 03
Motor-imagery classification and feature-selection workflow. Improving Classification of Multi-Class Motor Imagery by Statistical Feature Selection
Applied AI
Competition highlights
Practical experiments in segmentation, classification, and visual representation learning.
Cell Counting Competition
Shared top rank · 7 teams
Ovarian Cancer Subtype Classification & Outlier Detection
Top 20% · Rank 255 / 1,326
HuBMAP — Hacking the Human Vasculature
Top 21% · Rank 214 / 1,064
UW–Madison GI Tract Image Segmentation
Top 28% · Rank 421 / 1,565
Image Matching Challenge — CVPR Workshop
Top 24% · Rank 150 / 642
Notes & tutorials
Writing about how models learn.
Long-form explanations of papers, methods, and implementation choices in modern deep learning.
01
Large language models
DeepSeek-R1 → GRPO: Integrating Reinforcement Learning in LLMs
A practical reading of reward design, training stability, and Group Relative Policy Optimization.
Read article02
Representation learning
Self-Supervised Learning and a Case Study in SimCLR
Why learning from unlabeled images matters, and how contrastive learning creates useful representations.
Read article03
Transformers
Vision Transformer (ViT), Step by Step
A PyTorch-oriented walkthrough of patch embeddings, positional encodings, and the classification head.
Read article04
PyTorch
Module, Sequential, ModuleList, and ModuleDict
A guide to choosing the right PyTorch building block for clear, reusable model implementations.
Read articleResearch & collaboration
Let’s work on ambitious problems in visual AI.
I’m always happy to discuss research collaborations in machine learning, computer vision, and medical image processing.
m.dehghan9975@gmail.com