EE 7722 - What's New

Posted the take-home final, which is due Sunday, 5 May 2019. Please ask me for help if the questions are not clear.

The take-home final will be posted tomorrow. The first problem will be to apply interval analysis to the code from Homework 3 (one in which the mxv kernel does not load or store unneeded data). In the second problem a variation of UCNN (from Hegde 18) will be analyzed.

Homework 5 due time changed to Wednesday, 1 May 2019 at 16:00 CDT.

Homework 5 assigned, due Tuesday, 30 April 2019Wednesday, 1 April 2019 at 16:00 CDT.

Homework 4 assigned, due Monday, 29 April 2019. Additional problems will be assigned over the weekend (as Homework 5).

Start reading UCNN: exploiting computational reuse in deep neural networks via weight repetition. Access is free on campus.

Start reading this paper, Eyeriss: a spatial architecture for energy-efficient dataflow for convolutional neural networks. Access is free on campus.

The deadline for Homework 3 has been extended to Wednesday, 3 April 2019.

Start reading this paper, In-Datacenter Performance Analysis of a Tensor Processing Unit

Homework 3 assigned, due Monday, 1 Wednesday, 3 April 2019.

Homework 2 assigned, due Wednesday, 27 March 2019. This assignment is relatively easy. A more challenging homework will be assigned next, maybe over the weekend.

Linked a preliminary version of the interval analysis notes to the lecture notes page.

Homework 1 assigned, due Tuesday, 12 February 2019.

Linked the course overview slides to the lecture notes page.

Set up Web pages for the Spring 2019 semester.

Spring 2018 What's New.