Welcome to my website!
My name is Mehrdad Kiamari, and I am a Ph.D. student in the Department of Electrical and Computer Engineering (ECE) at the University of Southern California (USC). Currently, I am working under the guidance and supervision of Prof. Bhaskar Krishnamachari. My academic journey has been enriching, as I completed my Master's degree in Computer Science from USC in 2020. Before that, I pursued my Master's degree in Electrical Engineering from Sharif University of Technology and my Bachelor's degree in Electrical Engineering from the University of Tehran. As far as my research, I am focusing on the applications of deep neural networks, such as Graph Neural Networks and Reinforcement Learning, to enhance the performance of distributed computing networks. My recent work on applying Graph Neural Networks to improve scheduling has received a few awards (see my "Selected Research Projects" page).
Feel free to contact me at [last_name]@usc.edu
Educations:
Ph.D. in. Electrical Engineering at the University of Southern California (current program)
Master in Computer Science at the University of Southern California
Master in Electrical Engineering at Sharif University of Technology
Bachelor in Electrical Engineering at the University of Tehran
Current Research Interest:
Knowledge-graph Infused LLMs
Graph Neural Networks (GNN)
Reinforcement Learning (RL)
Federated Learning (FL)
Large Language Models (LLM)
Natural Language Processing (NLP)
Distributed Computing
Previous Research Interest:
Blockchain
Optimization for Network Problems
Covid-19 Risk Estimation through Mathematical and Data-Driven Modeling
Coded Distributed Computing
Achievements:
Technical Committee Member of the 2nd Graph Neural Networking (GNNet) Workshop 2023
Invited Paper at IEEE International Conference on Collaboration and Internet Computing (CIC) 2023
The 1st Place Winner of the 2nd Student Design Competition on Networked Computing on the Edge - 2022
2022 Best Poster - Honorable Mentioned by Ming Hsieh Department of Electrical and Computer Engineering
Winner-2020 International Covid-19 Computational Challenge-City of Los Angeles & RMDS Lab (licensed by USC Stevens Center for Innovation)
Nominee Best Paper Award and Best Student Paper Award - ACM Supercomputing Conference (SC’19)
Best Paper Award - IEEE Globecom Conference, 2017
Gold Medalist - Iranian nationwide electrical engineering Olympiad
Ranked 1st among all undergraduate electrical engineering students in the Telecommunication Department, University of Tehran
Selected Papers:
M. Kiamari and B. Krishnamachari, “Graph-Knowledge-Infused Vision-Language Models,” to be submitted to Journal of Machine Learning Research (JMLR) 2024
M. Kiamari, Y. Zhao, and B. Krishnamachari, “New low-cost auto-encoder Graph Neural Networks,” to be submitted to Journal of Machine Learning Research (JMLR) 2024
M. Kiamari, B. Krishnamachari, and S. Yun, “Blizzard: a Distributed Consensus Protocol for Mobile Devices," MDPI Journal on Mathematics, Feb. 2024
X. Li, M. Karakas, O. Hanna, M. Kiamari, J. Coleman, C. Fragouli, B. Krishnamachari and G. Verma, “Online Allocation of Sensing and Computation in Large Graphs,” invited paper at 9th IEEE International Conference on Collaboration and Internet Computing (CIC), Atlanta, USA, 2023
M. Kiamari and B. Krishnamachari ,"Gcnscheduler: Scheduling distributed computing applications using graph convolutional networks", Proceedings of the International Graph Neural Networking, December 2022, Rome, Italy
M. Kiamari, GS Ramachandran, Q Nguyen, E. Pereira, and J. Holm, B Krishnamachari “COVID-19 Risk Estimation using a Time-varying SIR-model,” Proceedings of the 1st ACM SIGSPATIAL International Workshop on Modeling and Understanding the Spread of COVID-19, Seattle WA USA, November 2020
M. Kiamari, B. Krishnamachari, M. Naveed, and S. Yun, “Distributed Consensus for Mobile Devices using Online Brokers,” in Proceedings of IEEE International Conference on Blockchain and Cryptocurrency (ICBC) 2020, Toronto, Canada, May 2020
K. Narra, Z. Lin, M. Kiamari, S. Avestimehr, M. Annavaram, “Distributed Computing Using Speed Adaptive Coding,” International Conference for High Performance Computing, Networking, Storage, and Analysis 2019 (nominated for both Best Student Paper and Best Paper finalist)
M. Kiamari, C. Wang, and S. Avestimehr, “On Heterogeneous Coded Distributed Computing,” in Proceedings of Globecom Conference 2017, December 2017 (Received Best Paper Award)
M. Kiamari, C. Wang, and S. Avestimehr, “SINR-Threshold Scheduling with Binary Power Control for D2D network,” in Proceedings of Globecom Conference 2017, December 2017
M. Kiamari, C. Wang, and S. Avestimehr, “Coding for Edge-Facilitated Wireless Distributed Computing with Heterogeneous Users,” Asilomar Conference on Signals, Systems, and Computers, October 2017
Work Experiences:
Graduate Research Assistant Present
University of Southern California Los Angeles, CA
Internship May. 2016 – Aug. 2016
Docomo Innovations, Inc. Palo Alto, CA
Professional Services:
IEEE Transaction on Information Theory
IEEE Transaction on Communications
IEEE Transaction on Networking
IEEE Globecom Conference
IEEE International Symposium on Information Theory
Teaching Assistant Experiences:
EE503: Probability for Electrical Engineering, USC
EE559: Pattern Recognition, USC
EE660: Machine Learning from Signal, USC
EE559: Pattern Recognition, USC
EE512: Stochastic Processes for Financial Engineering, USC
EE512: Stochastic Processes for Financial Engineering, USC
Selected Graduate Courses:
Fundamental Concepts of Analysis
Database Systems
Introduction to Mathematical Statistics
Deep Learning and its Applications
Stochastic Calculus and Mathematical Finance
Applied Natural Language Processing
Large-Scale Optimization for Machine Learning
Machine Learning
Mathematical Pattern Recognition
High-Dimensional Statistics and Big Data Problems
Foundation of Artificial Intelligence
Analysis of the Algorithm
Information Theory and Compression
Random Processes in Engineering
Probability for Electrical Engineering
Numerical Optimization
Advanced Communications
Selected Course Projects:
Citation Prediction for NeurIPS Papers using Natural Language Processing (NLP)
Collecting the title, authors & their bios, and abstract of all the published NeurIPS papers
Applied NLP algorithms (spaCy for NER, NLTK, TextBlob, RAKE for extracting keywords) to extract authors & their affiliations, keywords of the title and abstract of the papers as features for a classification task
Used a variety of models from the Scikit-learn package to build classifiers with over 90% accuracy
Memory-Efficient Convolutional Neural Network (CNN) Acceleration
Implemented ADMM-based optimization methods for both unstructured and structured pruning on convolutional neural networks, such as LeNet-5, AlexNet, and VGG16. Utilized different compressing schemes to encode the pruned model (sparse weight matrices) to reduce the amount of data that must be moved throughout the memory hierarchy
Deep Reinforcement Learning (Deep RL) with Tensorflow
Implemented the REINFORCE algorithm: rollout storage, policy network and training loop
Tested RL on various environments with engineering reward function as well as implementing Actor-Critic architecture
Deep Neural Network (Deep NN) with Tensorflow
Implemented the forward and backward passes as well as the NN training procedure for CNNs, RNNs, and GANs
Presentation:
GCNScheduler (the first GCN-based scheduling scheme), Research Festival at Ming Hsieh Department of Electrical and Computer Engineering - Received 2022 Best Poster - Honorable Mentioned