Alternative Computing Technologies (ACT) Lab

School of Computer Science

Georgia Institute of Technology

Tabla is an accelerator generator framework for geometric machine learning algorithms. It is an open source project under the Apache license.


Overview

TABLA is an innovative framework that generates accelerators for a class of machine learning algorithms. TABLA a template-based solution – from circuit to programming model – for using FPGAs to accelerate geometric machine learning algorithms. The objective of our solution is to devise the necessary programming abstractions and automated frameworks that are uniform across a range of machine learning algorithms. TABLA aims to avoid exposing software developers to the details of hardware design by leveraging commonalities in learning algorithms.

View details »



Citation

If you use this work, kindly cite our paper published in 22nd Annual IEEE International Symposium on High Performance Computer Architecture, 2016

D. Mahajan, J. Park, E. Amaro, H. Sharma, A. Yazdanbakhsh, J. Kim, H. Esmaeilzadeh, "TABLA: A Unified Template-based Framework for Accelerating Statistical Machine Learning

Benchmark List

TABLA supports a wide range of geometric machine learning algorithms. The following are the benchmark models we used for the experimental results in our paper:

Benchmarks
# Benchmark Name Algorithm Name Description Input Vectors # of Features Model Topology Lines of Code
1 LogisticR Logistic Regression Estimates the probability of dependent variable given one or more independent variables 581,000 54 54 20
2 LogisticR Logistic Regression Estimates the probability of dependent variable given one or more independent variables 500,000 200 200 20
3 SVM Classification Classifies data into different categoreis by identifying support vectors 581,000 54 54 23
4 SVM Classification Classifies data into different categoreis by identifying support vectors 500,000 200 200 23
5 Reco Recommender Systems Information filtering system that predict the preference a user would give to an item 1,700,000 27,000 1700x1000 31
6 Reco Recommender Systems Information filtering system that predict the preference a user would give to an item 24,000,000 100,000 6000x4000 31
7 Backprop Backpropagation Trains a neural network that model the mapping between the inputs and outputs of the data 38,000 10 10->9->1 48
8 Backprop Backpropagation Trains a neural network that model the mapping between the inputs and outputs of the data 90,000 256 256->128->256 48
9 LinearR Linear Regression Models the relationship between a dependent variable and one or more explanatory variables 10,000 55 55 17
10 LinearR Linear Regression Models the relationship between a dependent variable and one or more explanatory variables 10,000 784 784 17

License

This source code is published under the terms specified in the Apache license.

Copyright 2016 Hadi Esmaeilzadeh

Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at

http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.


Maintained by

Divya Mahajan


Contact

Please forward all your inquiries to: Divya, Hardik or Joon


Patrons