This repository contains the benchmarks for the IWLS 2020 programming contest.
The contest goal was to learn a Boolean function from a training set consisting of a small sample of random minterms of the function. The solutions were evaluated on a held-out set of minterms. The teams had the goal of maximizing the accuracy of the model while staying within a limit of 5,000 AIG nodes. The functions are provided in Espresso PLA format and the resulting models should be delivered in binary AIGER format. The instructions provided to the participants are here.
The benchmarks are also available in this repository and is composed of 100 functions in PLA, as follows:
Function ID | Function Category |
---|---|
00-09 | 2 MSBs for k-bits adders, with k in {16, 32, 64, 128, 256} |
10-19 | MSB of k-bit dividers and remainder circuits for k in {16, 32, 64, 128, 256} |
20-29 | MSB and middle bit of k-bit multipliers for k in {8, 16, 32, 64, 128} |
30-39 | k-bit comparators for k in {10, 20, ..., 100} |
40-49 | LSB and middle bit of k-bit square-rooters with k in {16, 32, 64, 128, 256} |
50-59 | 10 outputs of PicoJava designs with 16-200 inputs and roughly balanced on- & offset |
60-69 | 10 outputs of MCNC i10 design with 16-200 inputs and roughly balanced on- & offset |
70-79 | 5 other outputs from MCNC benchmarks + 5 symmetric functions of 16 inputs |
80-89 | 10 binary classification problems from MNIST group comparison |
90-99 | 10 binary classification problems from CIFAR-10 group comparison |
For further details on the contest including details on the benchmarks as well as the solutions of the different teams and their performance, please refer to our arXiv paper.