2.2.3 Part Three: Introduction to the final project – GraphChallenge, about 15
GraphChallenge is organized by MIT and Amazon, it encourages community approaches to
developing new solutions for analyzing graphs and sparse data derived from social media, sensor
feeds, and scientific data to enable relationships between events to be discovered as they unfold
Copyright By PowCoder代写 加微信 powcoder
in the field.
Sparse Deep Neural Network Graph Challenge is the newest challenge published in 2019.
This challenge performs neural network inference on a variety of sparse deep neural networks.
You can find slides, paper. example serial code, and example data sets from https:
//graphchallenge.mit.edu/challenges.
Our final project is a more specific version of the Sparse Deep Neural Network Graph
Challenge. The dataset and serial code will be introduced.
Requirements of the project
• Work individually, ask your TA for technical supports;
• Use sparsity of the matrices, and MPI+CUDA to program;
• Optimizations (data structure and algorithm design, sparsifying. MPI, CUDA, report writ-
ing, etc.) should be carefully considered;
All tests will be done on our 4-node 32-GPU cluster:
• Hand in your code and a final report describing your algorithm and performance.
Deadlines of the project
• Ist Deadline is Week 5
• Use slides to show your plan and your progress
• All students should come to the stage to present
No more than 1.5 minutes
• 2nd Deadline is Week 8
• Use slides to show your final performance results
• All students should come to the stage to show your achievements
• No more than 1.5 minutes
Hand in your code and a final report
程序代写 CS代考 加微信: powcoder QQ: 1823890830 Email: powcoder@163.com