Training Deep Neural Networks with Constrained Learning Parameters

被引:0
作者
Date, Prasanna [1 ]
Carothers, Christopher D. [1 ]
Mitchell, John E. [2 ]
Hendler, James A. [1 ]
Magdon-Ismail, Malik [1 ]
机构
[1] Rensselaer Polytech Inst, Dept Comp Sci, Troy, NY 12180 USA
[2] Rensselaer Polytech Inst, Dept Math Sci, Troy, NY 12180 USA
来源
2020 INTERNATIONAL CONFERENCE ON REBOOTING COMPUTING (ICRC 2020) | 2020年
关键词
Deep Neural Networks; Training Algorithm; Deep Learning; Machine Learning; Artificial Intelligence; LOCAL SEARCH; OPTIMIZATION;
D O I
10.1109/ICRC2020.2020.00018
中图分类号
TP301 [理论、方法];
学科分类号
081202 ;
摘要
Today's deep learning models are primarily trained on CPUs and GPUs. Although these models tend to have low error, they consume high power and utilize large amount of memory owing to double precision floating point learning parameters. Beyond the Moore's law, a significant portion of deep learning tasks would run on edge computing systems, which will form an indispensable part of the entire computation fabric. Subsequently, training deep learning models for such systems will have to be tailored and adopted to generate models that have the following desirable characteristics: low error, low memory, and low power. We believe that deep neural networks (DNNs), where learning parameters are constrained to have a set of finite discrete values, running on neuromorphic computing systems would be instrumental for intelligent edge computing systems having these desirable characteristics. To this extent, we propose the Combinatorial Neural Network Training Algorithm (CoNNTrA), that leverages a coordinate gradient descent-based approach for training deep learning models with finite discrete learning parameters. Next, we elaborate on the theoretical underpinnings and evaluate the computational complexity of CoNNTrA. As a proof of concept, we use CoNNTrA to train deep learning models with ternary learning parameters on the MNIST, Iris and ImageNet data sets and compare their performance to the same models trained using Backpropagation. We use following performance metrics for the comparison: (i) Training error; (ii) Validation error; (iii) Memory usage; and (iv) Training time. Our results indicate that CoNNTrA models use 32x less memory and have errors at par with the Backpropagation models.
引用
收藏
页码:107 / 115
页数:9
相关论文
共 32 条
  • [1] Abadi M, 2016, PROCEEDINGS OF OSDI'16: 12TH USENIX SYMPOSIUM ON OPERATING SYSTEMS DESIGN AND IMPLEMENTATION, P265
  • [2] Neural Algorithms and Computing Beyond Moore's Law
    Aimone, James B.
    [J]. COMMUNICATIONS OF THE ACM, 2019, 62 (04) : 110 - 119
  • [3] YodaNN1 : An Ultra-Low Power Convolutional Neural Network Accelerator Based on Binary Weights
    Andri, Renzo
    Cavigelli, Lukas
    Rossi, Davide
    Benini, Luca
    [J]. 2016 IEEE COMPUTER SOCIETY ANNUAL SYMPOSIUM ON VLSI (ISVLSI), 2016, : 236 - 241
  • [4] FINN-R: An End-to-End Deep-Learning Framework for Fast Exploration of Quantized Neural Networks
    Blott, Michaela
    Preusser, Thomas B.
    Fraser, Nicholas J.
    Gambardella, Giulio
    O'Brien, Kenneth
    Umuroglu, Yaman
    Leeser, Miriam
    Vissers, Kees
    [J]. ACM TRANSACTIONS ON RECONFIGURABLE TECHNOLOGY AND SYSTEMS, 2018, 11 (03)
  • [5] Local search heuristics for Quadratic Unconstrained Binary Optimization (QUBO)
    Boros, Endre
    Hammer, Peter L.
    Tavares, Gabriel
    [J]. JOURNAL OF HEURISTICS, 2007, 13 (02) : 99 - 132
  • [6] GCNet: Non-local Networks Meet Squeeze-Excitation Networks and Beyond
    Cao, Yue
    Xu, Jiarui
    Lin, Stephen
    Wei, Fangyun
    Hu, Han
    [J]. 2019 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION WORKSHOPS (ICCVW), 2019, : 1971 - 1980
  • [7] Choi J., 2018, ARXIV PREPRINT ARXIV
  • [8] Courbariaux M, 2015, ADV NEUR IN, V28
  • [9] Date P., 2019, THESIS RENSSELAER PO
  • [10] Date P, 2018, 2018 IEEE SYMPOSIUM SERIES ON COMPUTATIONAL INTELLIGENCE (IEEE SSCI), P242, DOI 10.1109/SSCI.2018.8628946