Hi, I'm YasminSalehi,

About Me

Yasmin Salehi

Hello and welcome to my personal website! My name is Yasmin, and I am currently working as a machine learning (ML) researcher at McGill University under the supervision of Prof. Giannacopoulos.

My current research involves developing novel physics-informed ML algorithms. Generally, my interests lie in graph representation learning, natural language processing (NLP), physics-informed ML, and application of ML in healthcare. I have recently published my work as the first author at Neural Information Processing Systems (NeurIPS2022) and Empirical Methods in Natural Language Processing (EMNLP2020) conferences.

I have previously earned a bachelor's and a master's degree in Electrical and Computer Engineering as well as two minors in Biomedical and Software Engineering from McGill University.



Skills

Python
Java
Matlab
C
SQL
C#
Assembly
PyTorch
PyTorch Geometric
NumPy
Scikit-learn
TensorFlow
TensorBoard
Pandas
NetworkX
NLTK
SeqIO
Git
Bash
Scrum
PyCharm
Google Colab
Google Cloud
BigQuery
HuggingFace
VS Code
CUDA

Education & Experience

  • McGill University Jan 2022 - Present
    Research Assistant
    PyTorch PyTorch Geometric Research Physics-Based Machine Learning
  • McGill University Jun 2021 - Aug 2021
    Research Assistant
    PyTorch PyTorch Geometric Research Physics-Based Machine Learning
  • McGill University Sept 2018 - Apr 2021
    M.Eng. Electrical and Computer Engineering
  • McGill University Sept 2012 - Apr 2018
    B.Eng. Electrical Engineering, Minor Software Engineering, Minor Biomedical Engineering

Publications

PhysGNN: A Physics-Driven Graph Neural Network Based Model for Predicting Soft Tissue Deformation in Image-Guided Neurosurgery

NeurIPS2022

Yasmin Salehi, Dennis Giannacopoulos

TLDR: PhysGNN is a novel physics-driven graph neural network based model capable of accurately and efficiently approximating tissue deformation caused by applied forces by which 94-97% of errors are shown to be less than 1 mm—the precision in neurosurgery.

Publication
Graph Neural Networks (GNNs) Physics-based simulation Finite Element Method (FEM)

Structure Aware Negative Sampling (SANS) in Knowledge Graphs

EMNLP 2020

Kian Ahrabian*, Aarash Feizi*, Yasmin Salehi*, William Hamilton, Joey Bose

TLDR: SANS is a novel negative sampling method which generates negative samples from the k-hop neighborhood of nodes. SANS was shown to be competitive with SOTA methods and outperform sophisticated Generative Adversarial Network approaches.

Publication
Negative Sampling Knowledge Graphs

Contact