Hi, I'm Hooman Zolfaghari

Computer Science Enthusiast | Passionate About Probability & Machine Learning Research

About Me

I’m a Computer Science student at Sharif University of Technology, with a deep passion for Probability and Machine Learning . I’m always searching for the math and reasoning behind everything, which drives my focus into the mathematical concepts of Machine Learning.
I naturally do a lot of programming as well, you may see some of my public codes on my GitHub.
Recently, I’ve had the opportunity to work on exciting research projects. you can see my research experience here.

Here you can download a concise version of my academic resume.

Research Interests

Understand at a precise mathematical level what capabilities and information are fundamentally needed to learn different kinds of tasks successfully, and to understand the basic algorithmic principles involved in getting computers to learn from data and to improve performance with feedback (ref).
Proving guarantees for algorithms (under what conditions will they succeed ?, how much data and computation time is needed ?) and developing machine learning algorithms that provably meet desired criteria.
Particularly researching problems in Sample Comlexity, Learnability, Generalization, Stability, Kernels, or more specific problems like secnarios like Deep Learning Theory, Domain Adaptation, Mathematics of Attention, Nueral Tangent Kernel
Consquently, I deeply enjoy topics related to High-Dimensional Probability and Statistics.

geometric unification of a broad class of ML problems from the perspectives of symmetry and invariance (ref). Extending models and theories to use non-Euclidean data, such as graphs. Learn and generalize from structured data by incorporating geometric priors into the models.

Creating secure and private models without losing perfomance. Out-of-Distribution Generalization, Explainability, Uncertainty Quantification, and Evaluation of Trustworthiness (ref)
Particular examples: Differential Privacy, Certfied Robustness, Distributional Robustness, Catastrophic Overfitting, Memebrship Inference Attacks and Adversarial Examples, Machine Unlearning.

I am also highly curios and motivated for research and gaining more experience in these areas

Generative Deep Learning

Learning complex and high-dimensional data distributions, Universal Distribution Approximation, Text Guided Generation or Editing

Reinforcement Learning

Descision making under Uncertainty, Multi-Agent RL, Inverse RL, Bandits

Causal Machine Learning

Shifting the focus of learning from finding correlations to discovering causality, Learning Causal Structure Representations

In addition to my research interests, I find the following topics interesting to study:

Randomized Algorithms and Sampling Algorithms
Computational Neuroscience
Quantum Machine Learning
Math topics in Topology, Group Theory, Measure Theory, Optimization

Working on Neural Tangent Kernel and other Neural Networks Theories in my High-Dimensional Probability course project.

Other than my theoretical studies I spent a lot of time coding and experimenting. I have worked with many languages and frameworks. I have worked with ML libraries in advanced levels. For example Pytorch, Pytorch Geometric, CVXPY and CVXPy Layers, Gymnasium and etc.

Attended Neuromatch Academy's Deep Learning Course (Certificate)

Studied CS236: Deep Generative Models, Stanford

Studied CS224W: Machine Learning with Graphs, Stanford

Implemented various machine learning models and techniques (GitHub)

Stochastic Processes course Project, MCMC:Metropolis-Hastings algorithm (GitHub)

Education

B.S in Computer Science at Sharif University of Technology

Courses of Interest:

Machine Learning

This was a graduate course since it contained some topics of ML Theory as well. It was my first official ML course where I learned some of the theoretical foundations and topics ranging from classics like random forests and mixture models to Neural Networks and RL. And I implemented some methods from scratch, like MAP, SVM and MLP. I got 19.3 out of 20. The final project was also Nueral Style Transfer.

Stochastic Processes

This was my first advanced probability course, I got 20 out of 20. We learned about Markov Chains and its variations, Poisson and Gaussian processes, martingals and many other topics. The final project and my extra project was about MCMC: Metropolis Hastings algorithm, where I was introduced to sampling methods.

High-Dimensional Probability Analysis

This is a graduate course where we study probabilistic methods and concentration inequalities in high-dimensional spaces, with applications in areas such as machine learning, random matrix theory, and statistics. I am currently studying this course.

Theory of Machine Learning

This is a graduate course where we learn many of the mathematical concepts of today's Machine Learning and Deep Learning. I am currently studying this course.

Statistical Learning

In this course I learned a formal statistical learning view that is incorporated in ML. This was an introductory course compared to the last ones I mentioned.

Linear Algebra, Game Theory

These courses were also among my favorite courses.

Probability and its applications, Statistics and its applications

These are also foundation courses that I enjoyed.

Databases, Data Structures and Algorithms, Advanced Programming, Operating Systems, Computer Networks, Principles of Computer Systems, Numerical Analysis, Automata and Theory of languages

These courses also helped me through my CS journy and are foundations of my Computer Science knowledge.

Other Certificates:

Deep Learning from Neuromatch Academy

“You keep on learning and learning, and pretty soon you learn something no one has learned before.”

(Probably) Richard Feynman

Experience

Research Experience

Graph Similarity Search, under the supervision of Dr. A. Rafiey and Dr. A. Vaezi.
We worked on Expressive measures of graph similarity for ML tasks. Starting with non-parametric and kernel methods, and then incorporating deep parametric methods. Here I had the opportunity to introduce many ideas and theoretical analysis, create theoretical proofs and conduct benchmark experiments, under the guidance of my supervisors. I studied concepts in Geometric Deep Learning and Graph NNs, Approximation on Graphs, Graph Theory, Optimization and Topology. Paper is in draft stage.
On Security and Privacy of Kolmogorov-Arnold Networks, under the supervision of Prof. S. Amini:
We are experimenting with many aspects of Trustworthy ML on KANs, such as Adversarial Robustness, Membership Inference, Certified Robustness, Catastrophic overfitting, and Machine Unlearning. Collaborated with students at the University of Massachusetts
Bayes-Markov model of Cortical Orientation Selectivity, under the supervision of Dr. H. Peyvandi.
We studied and experimented on a model based on Markov Random Fields and Bayesian Inference, inspired by their use in vision tasks.

Teaching Experience

This is a graduate course at the Electrical Electrical Engineering department at Sharif University of Technology. I am currently responsible for theoretical and practical exercise on Generative Flow Models. I also mentor students introducing them to related research areas and recent papers. Course Page

This course is for both Undergraduate and Graduate students at SUT, it was also available internationally to be audited. I am currently part of the team for educational content of “Breakthroughs: Vision Transformers, Self-Supervised Learning, Contrastive Learning” Chapter and the team for Unsupervised Learning and Clustering where I created the course slides. Course Page

In this course I was responsible for holding mentorship classes every week, covering topics such as python programming, combinatorics, sorting algorithms, induction and greedy algorithms, game theory and combinatorial games, graph theory and graph algorithms, recursion and recursive programming, and dynamic programming.

In this course, I was responsible for designing and scoring students final projects. I also held workshops guiding studets to create a custom application layer protocol and use TCP and UDP to create client-server and P2P network paradigms with socket programming. Workshop codes and videos are here.

I created exercises in five series throughout the course, and deployed them on Quera for students. My questions included the topics of dynamic programming, red-black trees, hash-tables, sorting etc.

In this course, I created and scored exercises on OS processes.

In this course I held TA classes on various topics including Git, File storage, Clean Code, OOP etc. I also created educational content on Multi Threaded programming and Reflection in Java. And I helped with scoring student exercises.

Industrial Experience

Software Development, Mika Corporation

Contact

Feel free to reach out to me at hoomanzolfaghari84@gmail.com or hooman.zolfaghari84@sharif.edu.