About

Mohammad Alizadeh is the NEC Professor of Software Science and Engineering at MIT, serving as an Industry Officer and Director of the 6-A MEng Thesis Program. His research focuses on developing groundbreaking systems in the fields of electrical engineering and computer science, particularly in areas related to software engineering, systems, and networking. As a faculty member, he contributes to advancing knowledge in these domains through his leadership and research activities.

Research topics

• Computer Science
• Information Retrieval
• Computer network
• Real-time computing
• Telecommunications
• Operating system
• Data Mining
• Computer Security
• Distributed computing
• Programming language
• World Wide Web
• Materials science
• Engineering
• Embedded system

Selected publications

• Theoretical error modeling and analysis of strapdown inertial navigation system alignment under zero-velocity conditions

  Scientific Reports · 2025-11-21 · 1 citations

  articleOpen accessSenior author

  One of the important stages before the operation of an inertial navigation system is the initial alignment procedure, known as fine alignment under stationary conditions. This study analyzes the accuracy of the fine alignment algorithm, which uses zero-velocity updates to estimate and compensate navigation errors. Since the velocity-matching approach is not fully observable, an observability analysis identifies which error states cannot be estimated. The unobservable states and thus the ultimate alignment accuracy depend on both sensor precision and the orientation of the inertial measurement unit. This paper investigates the influence of attitude on fine alignment performance through analytical derivations and numerical simulations. New relationships are developed to predict alignment accuracy for arbitrary orientations, and their validity is demonstrated via extensive simulation results. Additionally, the impact of x-, y-, and z-channel sensors on the final accuracy of inertial alignment is examined statistically. The results reveal a differential sensitivity, indicating that alignment accuracy is predominantly influenced by sensors along the z-channel. This insight provides a practical strategy for sensor placement optimization, suggesting that allocating higher-accuracy sensors to the z-axis within a fixed budget can yield significant improvements in overall navigation system performance.

  PublisherOA PDFDOI

• Urban building energy modelling for retrofitting at scale: state-of-the-art and future prospects

  IOP Conference Series Earth and Environmental Science · 2025-11-01

  articleOpen access1st authorCorresponding

  Abstract Energy retrofitting of existing buildings is an effective strategy for mitigating climate change, offering co-benefits such as improved quality of life, health, and economic growth. However, scaling retrofitting efforts to meet climate targets remains challenging, as it traditionally requires costly expertise to identify high-potential buildings and optimal energy conservation measures (ECMs). Urban Building Energy Modelling (UBEM) has emerged as a powerful tool for large-scale assessments, evaluating ECM effects across districts and cities. Yet, effectively applying retrofitting UBEM (RUBEM) requires careful attention to model complexity, data availability, and the context-specific purpose of modelling. Current automated tools, while helpful, require extensive data and simulations, and still yield estimates with high uncertainty, highlighting the need for new scalable models. While various RUBEMs have been reported in the literature, a thorough analysis is needed to identify challenges and the effective strategies to address them. This study aims to set the research agenda for RUBEM. It presents a review of existing RUBEMs, looking into their core objectives, capabilities, and performance to define functional requirements, challenges, and modelling. The review has shown that typical model applications include matching ECMs to buildings, estimation of ECM effects on individual and system levels, and optimisation of renovation strategies across large portfolios with respect to economic, environmental, energy, cultural, and construction constraints. Data scarcity and quality, model accuracy and validity, model complexity, and computational burden were identified as challenges. These lead to trade-offs between accuracy, speed, and feasibility, which should be driven by the purpose and scale of each RUBEM.

  PublisherDOI

• Automated drug design for druggable target identification using integrated stacked autoencoder and hierarchically self-adaptive optimization

  Scientific Reports · 2025-09-01 · 1 citations

  articleOpen accessSenior author

  Drug classification and target identification are crucial yet challenging steps in drug discovery. Existing methods often suffer from inefficiencies, overfitting, and limited scalability. Traditional approaches like support vector machines and XGBoost struggle to handle large, complex pharmaceutical datasets effectively. Deep learning models, while powerful, face challenges with interpretability, computational complexity, and generalization to unseen data. This study addresses these limitations by introducing a novel framework: optSAE + HSAPSO. This framework integrates a stacked autoencoder (SAE) for robust feature extraction with a hierarchically self-adaptive particle swarm optimization (HSAPSO) algorithm for adaptive parameter optimization. This combination delivers superior performance across various classification metrics. Experimental evaluations on datasets from DrugBank and Swiss-Prot demonstrate that optSAE + HSAPSO achieves a high accuracy of 95.52%. Notably, it exhibits significantly reduced computational complexity (0.010 s per sample) and exceptional stability (± 0.003). Compared to state-of-the-art methods, the framework offers higher accuracy, faster convergence, and greater resilience to variability. Furthermore, ROC and convergence analyses confirm its robustness and generalization capability, maintaining consistent performance across both validation and unseen datasets. By leveraging advanced optimization techniques, the framework efficiently handles large feature sets and diverse pharmaceutical data, making it a scalable and adaptable solution for real-world drug discovery applications. However, the method's performance is dependent on the quality of the training data, and fine-tuning may be necessary for high-dimensional datasets. Despite these limitations, the optSAE + HSAPSO framework demonstrates transformative potential, significantly reducing computational overhead while improving classification accuracy and reliability. This work advances the field of pharmaceutical informatics by presenting a reliable and efficient framework for drug classification and target identification. These findings open promising avenues for future research, including extending the framework to other domains such as disease diagnostics or genetic data classification, ultimately accelerating the drug development process.

  PublisherOA PDFDOI

• <scp>GraphPipe:</scp> Improving Performance and Scalability of DNN Training with Graph Pipeline Parallelism

  2025-02-06 · 5 citations

  articleOpen access

  Deep neural networks (DNNs) continue to grow rapidly in size, making them infeasible to train on a single device (e.g. GPU). Pipeline parallelism is commonly used in existing DNN systems to support large-scale DNN training by partitioning a DNN into multiple stages, which concurrently perform DNN computation for different micro-batches of training samples in a pipeline fashion. However, existing pipeline-parallel approaches only consider sequential pipeline stages and thus ignore the topology of a DNN, resulting in missed model-parallel opportunities.

  PublisherDOI

• Analytical quaternion-based bias estimation algorithm for fast and accurate stationary gyro-compassing

  Scientific Reports · 2024-07-09 · 9 citations

  articleOpen accessSenior author

  This work introduces a novel approach to Strapdown Inertial Navigation System (SINS) alignment, distinct from recursive methods like Kalman filtering. The proposed methodology expedites bias error calculations by utilizing quaternion-based analytical relationships, which bypasses the slow convergence behavior associated with recursive algorithms, particularly in azimuth angle error estimation. In addition, the proposed approach demonstrates comparable accuracy to traditional fine alignment methods. Simulations and experiments validate that in contrast to the 10-min time requirement of traditional fine alignment methods (for azimuth angle estimation in stationary conditions), the proposed approach achieves the same accuracy within 20 s. However, limitations exist as the algorithm is applicable only in stationary conditions, and necessitating a high-grade IMU capable of measuring the earth's rotation rate.

  PublisherOA PDFDOI

• Principles for Internet Congestion Management

  2024-07-31 · 4 citations

  articleOpen access

  Given the technical flaws with---and the increasing non-observance of---the TCP-friendliness paradigm, we must rethink how the Internet should manage bandwidth allocation. We explore this question from first principles, but remain within the constraints of the Internet's current architecture and commercial arrangements. We propose a new framework, Recursive Congestion Shares (RCS), that provides bandwidth allocations independent of which congestion control algorithms flows use but consistent with the Internet's economics. We show that RCS achieves this goal using game-theoretic calculations and simulations as well as network emulation.

  PublisherOA PDFDOI

• m3: Accurate Flow-Level Performance Estimation using Machine Learning

  2024-07-31 · 8 citations

  articleOpen access

  Data center network operators often need accurate estimates of aggregate network performance. Unfortunately, existing methods for estimating aggregate network statistics are either inaccurate or too slow to be practical at the data center scale.

  PublisherDOI

• Deep Reinforcement Learning Based Network Reconfiguration

  2023-12-05 · 1 citations

  article1st authorCorresponding

  Implementing energy management (EM) programs in modern power systems is pivotal in optimizing energy consumption, enhancing grid stability, and reducing operational costs. However, the success of these EM programs is often influenced by the rebound effect, where the electricity load returns to its previous high levels after an initial reduction, mitigating the intended energy-saving benefits. Addressing this challenge, this research proposes a novel approach for network reconfiguration (NRC) using Deep Reinforcement Learning (DRL) with a Dueling Deep Q-Network (DDQN) algorithm. Our methodology leverages DDQN to dynamically adapt the network configuration to changing electricity load patterns influenced by EM programs. The NRC process is formulated as a DRL problem, where the agent learns to make optimal decisions regarding minimizing operational costs. Studies results on the IEEE 33 bus system demonstrate the effectiveness of the proposed DDQN-based NRC approach in mitigating the destructive impacts of the rebound effect.

  PublisherDOI

• Bringing Reconfigurability to the Network Stack

  arXiv (Cornell University) · 2023-11-13

  preprintOpen access

  Reconfiguring the network stack allows applications to specialize the implementations of communication libraries depending on where they run, the requests they serve, and the performance they need to provide. Specializing applications in this way is challenging because developers need to choose the libraries they use when writing a program and cannot easily change them at runtime. This paper introduces Bertha, which allows these choices to be changed at runtime without limiting developer flexibility in the choice of network and communication functions. Bertha allows applications to safely use optimized communication primitives (including ones with deployment limitations) without limiting deployability. Our evaluation shows cases where this results in 16x higher throughput and 63% lower latency than current portable approaches while imposing minimal overheads when compared to a hand-optimized versions that use deployment-specific communication primitives.

  PublisherOA PDFDOI

• Tight Loops, Smooth Streams: Responsive Congestion Control for Real-Time Video

  arXiv (Cornell University) · 2023-09-28

  preprintOpen accessSenior author

  Real-time video streaming relies on rate control mechanisms to adapt video bitrate to network capacity while maintaining high utilization and low delay. However, the current video rate controllers, such as Google Congestion Control (GCC), are very slow to respond to network changes, leading to link under-utilization and latency spikes. While recent delay-based congestion control algorithms promise high efficiency and rapid adaptation to variable conditions, low-latency video applications have been unable to adopt these schemes due to the intertwined relationship between video encoders and rate control in current systems. This paper introduces Vidaptive, a new rate control mechanism designed for low-latency video applications. Vidaptive decouples packet transmission decisions from encoder output, injecting ``dummy'' padding traffic as needed to treat video streams akin to backlogged flows controlled by a delay-based congestion controller. Vidaptive then adapts the target bitrate of the encoder based on delay measurements to align the video bitrate with the congestion controller's sending rate. Our evaluations atop Google's implementation of WebRTC show that, across a set of cellular traces, Vidaptive achieves ~1.5x higher video bitrate and 1.4 dB higher SSIM, 1.3 dB higher PSNR, and 40% higher VMAF, and it reduces 95th-percentile frame latency by 2.2 s with a slight 17 ms increase in median frame latency.

  PublisherOA PDFDOI

Recent grants

• Collaborative Research: CNS Core: Small: Understanding Per-Hop Flow Control

  NSF · $250k · 2020–2022

• CNS Core: Small: Network Architecture and Routing Protocols for Payment Channel Networks

  NSF · $500k · 2019–2022

• NeTS: Small: Collaborative Research: A Fast and Flexible Transport Architecture for High Speed Networks

  NSF · $250k · 2016–2019

• CAREER: Data-Driven Network Resource Management Systems

  NSF · $628k · 2018–2024

Frequent coauthors

• Hari Balakrishnan

  IIT@MIT

  25 shared

• Balaji Prabhakar

  Jawaharlal Nehru Technological University, Hyderabad

  23 shared

• Prateesh Goyal

  19 shared

• Scott Shenker

  University of California, Berkeley

  16 shared

• Ryan Marcus

  California University of Pennsylvania

  15 shared

• Changhoon Kim

  14 shared

• Tim Kraska

  Amazon (United States)

  14 shared

• Nick McKeown

  Aberystwyth University

  13 shared

Education

• Ph.D., Electrical Engineering and Computer Science

  Massachusetts Institute of Technology

  2008

• M.S., Electrical Engineering and Computer Science

  Massachusetts Institute of Technology

  2004

• B.S., Electrical Engineering

  Sharif University of Technology

  2002