Research

  • New Parallelism Models for Deep Learning Inference: Implementing a hybrid parallelism model that turns idle threads into computing reinforcement for straggler threads with zero-data movement.
  • New Parallelism Models for Graph Computing: Proposing the new MPI*X parallelism model that leverages thread-based partitioning and placement and deems threads as basic units of computing and communication.
  • Distributed Sparse Data Structures and Algorithms: Devising the new Triply Compressed Sparse Column (TCSC) format for Sparse Matrix – Sparse input and output Vectors (SpMSpV2) primitive.
Past research
  • Reinforcement Learning based Graph Partitioning: A distributed graph partitioning algorithm which leverages reinforcement learning and label propagation to form highly localized and balanced partitions.
  • Grid Resource Discovery: Performing intuitive and adaptive resource discovery operation on a fully distributed computational Grid by Distributed Learning Automata (DLA). The Grid overlay network is reconstructed by a network of distributed learning automata which gradually learn optimal routes for different kinds of resource queries.
  • Applied Machine Learning to Optimization Algorithms: Applying Learning Automata (LA) to Evolutionary Algorithms (EAs) such as Particle Swarm Optimization (PSO) and Group Search Optimizer (GSO) for improving their performance, speed, and robustness.
  • Applied Machine Learning to Image Processing: Utilizing Learning Automata (LA) for unsupervised and adaptive edge detection of binary and grayscale images.

Publications

  • 2020

  • Mohammad Hasanzadeh Mofrad. “Distributed Sparse Computing and Communication for Big Graph Analytics and Deep Learning.” University of Pittsburgh, 2020. full-text presentation
  • Mohammad Hasanzadeh Mofrad, Rami Melhem, Yousuf Ahmad and Mohammad Hammoud. “Graphite: A NUMA-aware HPC System for Graph Analytics Based on a new MPI * X Parallelism Model.” In proceedings of Very Large Data Bases (PVLDB), 13(6), 783-797, Tokyo, Japan, 2020. full-text presentation code
  • Mohammad Hasanzadeh Mofrad, Rami Melhem, Yousuf Ahmad and Mohammad Hammoud. “Accelerating Distributed Inference of Sparse Deep Neural Networks via Mitigating the Straggler Effect.” In proceedings of IEEE High Performance Extreme Computing (HPEC), Waltham, MA USA, 2020. full-text presentation code
  • Mohammad Hasanzadeh Mofrad, Rami Melhem, Yousuf Ahmad and Mohammad Hammoud. “Studying the Effects of Hashing of Sparse Deep Neural Networks on Data and Model Parallelisms.” In proceedings of IEEE High Performance Extreme Computing (HPEC), Waltham, MA USA, 2020. full-text presentation code

  • 2019

  • Mohammad Hasanzadeh Mofrad, Rami Melhem, Yousuf Ahmad and Mohammad Hammoud. “Efficient Distributed Graph Analytics using Triply Compressed Sparse Format.” In proceedings of IEEE Cluster Computing (CLUSTER), Albuquerque, NM USA, 2019 (Acceptance rate 27%). full-text presentation code
  • Mohammad Hasanzadeh Mofrad, Rami Melhem, Yousuf Ahmad and Mohammad Hammoud. “Multithreaded Layer-wise Training of Sparse Deep Neural Networks using Compressed Sparse Column.” In proceedings of IEEE High Performance Extreme Computing (HPEC), Waltham, MA USA, 2019. full-text code
  • Mohammad Hasanzadeh Mofrad, Rami Melhem, and Mohammad Hammoud. “Partitioning Graphs for the Cloud using Reinforcement Learning.” arXiv:1907.06768, 2019. full-text code

  • 2018

  • Mohammad Hasanzadeh Mofrad, Rami Melhem, and Mohammad Hammoud. “Revolver: Vertex-centric Graph Partitioning Using Reinforcement Learning.” In proceedings of IEEE CLOUD (CLOUD), San Francisco, CA USA, 2018 (Acceptance rate 23%). full-text presentation code
  • Mohammad Hasanzadeh Mofrad and Daniel Mosse. “Speech Recognition and Voice Separation for the Internet of Things.” In Proceedings of the 8th ACM International Conference on the Internet of Things (ACM IoT), Santa Barbara, CA USA, 2018 (Acceptance rate 34%). full-text presentation code
  • Mohammad Hasanzadeh Mofrad and S. K. Chang. “A Bi-population Particle Swarm Optimizer for Learning Automata based Slow Intelligent System.” arXiv:1804.00768, 2018. full-text presentation code

  • 2017

  • Mohammad Hasanzadeh Mofrad, and Alireza Rezvanian. "Learning Automata Clustering." Journal of Computational Science, 2017. full-text code
  • Mohammad Hasanzadeh Mofrad, and Adam Lee. "Leveraging Intel SGX to Create a Nondisclosure Cryptographic library." arXiv:1705.04706, 2017. full-text presentation code ...
    Intel Software Guard Extension (Intel SGX) is an Intel technology that puts sensitive code and data into CPU-hardened protected regions called enclaves. In this project we leverage the Intel SGX to produce a secure cryptographic library which keeps the generated keys inside an enclave restricting use and dissemination of confidential cryptographic keys. Using enclaves to store the keys we maintain a small Trusted Computing Base (TCB) where we also perform computation on temporary buffers to and from untrusted application code. As a proof of concept, we implemented hashes and symmetric encryption algorithms inside the enclave where we stored hashe values, Initialization Vectors (IVs) and random keys.
  • Debashis Ganguly, Mohammad Hasanzadeh Mofrad, Taieb Znati, Rami Melhem, and John R. Lange. "Harvesting Underutilized Resources to Improve Responsiveness and Tolerance to Crash and Silent Faults for Data-intensive Applications." In proceedings of 10th IEEE International Conference on Cloud Computing (CLOUD), Honolulu, Hawaii USA, 2017. full-text presentation code
  • Debashis Ganguly, Mohammad Hasanzadeh Mofrad, and Adriana Kovashka. "Detecting Sexually Provocative Images." In proceedings of IEEE Winter Conference on Applications of Computer Vision (WACV), Santa Rosa, CA USA, 2017. full-text presentation poster supplementary dataset ...
    Designing a hierarchical framework from scratch for predicting intentions behind images using machine learning. This project consists of three major phases. Creating a new dataset of celebrities persuasive intents by collecting 1146 images from 204 celebrities and crowdsourcing the dataset using the Amazon Mechanical Turk to identify the persuasive intents. After creating this dataset, we create a multilevel hierarchical framework using support vector machines to learn the persuasive intents from training images. Finally, a model is created which is capable of inferring persuasive intents from test images with a fair accuracy.

  • 2016

  • Mohammad Hasanzadeh Mofrad, Omid Jalilian, Alireza Rezvanian, and Mohammad Reza Meybodi. "Service Level Agreement based Adaptive Grid Superscheduling." Future Generation Computer Systems (FGCS), 55, 62-73, 2016. full-text code
  • Mohammad Hasanzadeh, Sana Sadeghi, Alireza Rezvanian, and Mohammad Reza Meybodi. "Success Rate Group Search Optimizer." Journal of Experimental & Theoretical Artificial Intelligence, 28.1-2, 53-69, 2016. full-text code

  • 2015

  • Mohammad Hasanzadeh Mofrad, Sana Sadeghi, Alireza Rezvanian, and Mohammad Reza Meybodi. "Cellular Edge Detection: Combining Cellular Automata and Cellular Learning Automata." AEÜ - International Journal of Electronics and Communications, 69.1, 1282-1290, 2015. full-text code
  • Mohammad Hasanzadeh and Mohammad Reza Meybodi. "Distributed Optimization Grid Resource Discovery." The Journal of Supercomputing, 71.1, 87-120, 2015. full-text code

  • 2014

  • Mohammad Hasanzadeh and Mohammad Reza Meybodi. "Grid Resource Discovery Based on Distributed Learning Automata" Computing, 96.2, 909-922, 2014. full-text code

  • Mohammad Hasanzadeh, Mohammad Reza Meybodi, and Mohammad Mehdi Ebadzadeh. "Adaptive Parameter Selection in Comprehensive Learning Particle Swarm Optimizer." In proceedings of 17th International Symposium on Artificial Intelligence and Signal Processing (AISP), 267–276, Tehran, Iran, 2014. full-text presentation code

  • Mohammad Hasanzadeh, Hossein Hamidi, and Habibollah Asghari. "Plaintext Transmission over Session Initiation Protocol." In proceedings of 7th International Symposium on Telecommunications (IST), 629-634, Tehran, Iran, 2014. full-text presentation
  • Mohammad Hasanzadeh, Mohammad Reza Meybodi, and Mohammad Mehdi Ebadzadeh. "A Learning Automata Approach to Cooperative Particle Swarm Optimizer." Journal of Information Systems and Telecommunication, 2.5, 1-14, 2014. full-text code

  • 2013

  • Mohammad Hasanzadeh, Mohammad Reza Meybodi, and Mohammad Mehdi Ebadzadeh "Adaptive Cooperative Particle Swarm Optimizer." Applied Intelligence, 39.2, 397-420, 2013. full-text code

  • Mohammad Hasanzadeh, Mohammad Reza Meybodi. "Deployment of gLite Middleware: An E-science Grid Infrastructure" In proceedings of 21st Iranian Conference on Electrical Engineering (ICEE), 1-6, Mashhad, Iran, 2013. full-text poster

  • 2012

  • Mohammad Hasanzadeh, Mohammad Reza Meybodi, and Mohammad Mehdi Ebadzadeh. "A Robust Heuristic Algorithm for Cooperative Particle Swarm Optimizer: A Learning Automata Approach." in proceedings of 20th Iranian Conference on Electrical Engineering (ICEE), 656-661, Tehran, Iran, 2012. full-text presentation code

  • 2011

  • Mohammad Hasanzadeh, Mohammad Reza Meybodi, Saeed Shiry Ghidary. "Improving Learning Automata based Particle Swarm: An optimization algorithm." In proceedings of 12th International Symposium on Computational Intelligence and Informatics (CINTI), 291-296, Budapest, Hungary, 2011. full-text code