SuperUROP All SuperUROP Projects (eecs)
spacing
SuperUROP
 
Home
Letters of Recommendation
Login

SuperUROP Projects

"AI is for Kids"
Faculty Advisor: Hal Abelson
Mentor(s):
Contact e-mail: hal@mit.edu
Research Area(s):
The MIT App Inventor group (appinventor.mit.edu) works to empower young people worldwide to create original applications with mobile technology. This work will extend the App Inventor platform to encompass the neural network features in the emerging generation of smartphones and create new project opportunities for kids to build AI applications.
"Faster DNNs in TensorFlow"
Faculty Advisor: Saman Amarasinghe
Mentor(s): Riyadh Baghdadi
Contact e-mail: baghdadi@mit.edu
Research Area(s): Artificial Intelligence, Computer Systems
The COMMIT group (Link is developing a new code optimization framework for DSL (Domain Specific Language) compilers called Tiramisu. It takes high level code generated by a DSL compiler, optimizes it and then generates highly optimized code targeting multiple hardware architectures such as multicores, GPUs, FPGAs and distributed clusters. The goal of this project is to integrate Tiramisu within TensorFlow and use it to generate faster DNNs.

We have already integrated Tiramisu with success in two DSL compilers: Halide and Julia. Tiramisu extended Halide and Julia with many new capabilities including the ability to express new algorithms, the ability to perform new complex loop nest transformations and the ability to generate efficient code for more architectures. It led to speedups reaching 4x in Halide and 16x in Julia due to optimizations enabled by Tiramisu.

Preliminary experiments in implemented DNN convolutions in Tiramisu showed that convolutions generated by Tiramisu are 1.5 times faster than convolutions implemented in the Intel MKL-DNN library (the most optimized DNN library for Intel architectures).

Requirements
Knowledge of TensorFlow and deep learning.
Knowledge of compilers.
Taking 6.172/6.871 (Performance Engineering of Software Systems) is plus.
"Parallel Streaming Graph Analytics"
Faculty Advisor: Julian Shun
Mentor(s):
Contact e-mail: jshun@mit.edu
Research Area(s): Computer Systems, Theoretical Computer Science
Graph algorithms have applications in a variety of domains, such as social network and Web analysis, computational biology, and machine learning. Analyzing large graphs quickly requires designing high-performance parallel algorithms. Furthermore, graphs are changing quickly, requiring data structures that allow efficient parallel updates and dynamic algorithms that respond to changes without recomputing from scratch. This project involves the design and implementation of parallel dynamic graph algorithms and possibly a programming framework that improves the programmability of these algorithms. This project is suitable for students who are interested in algorithm analysis and performance engineering.

Prerequisites: Performed well in 6.172 and 6.046 or have comparable experience. Committed to at least 15 hours/week.
"Theory and Practice of Parallel Algorithms"
Faculty Advisor: Julian Shun
Mentor(s):
Contact e-mail: jshun@mit.edu
Research Area(s): Computer Systems, Theoretical Computer Science
This project involves the design, analysis, and implementation of parallel algorithms for various fundamental problems in computing such as subgraph matching, graph connectivity, clustering, text analytics, etc. The goal is to design parallel algorithms that perform well both in theory and in practice. This project is suitable for students who are interested in algorithm analysis and performance engineering.

Prerequisites: Performed well in 6.172 and 6.046 or have comparable experience. Committed to at least 15 hours/week.

Total: 4

Technical questions related to this site? Please contact Search-Master@superurop-apply.mit.edu