We present LBW-Net, an efficient optimization based method for quantization and
training of the low bit-width convolutional neural networks (CNNs). Specifically, we quantize
the weights to zero or powers of 2 by minimizing the Euclidean distance between
full-precision weights and quantized weights during back-propagation (weight learning).
We characterize the combinatorial nature of the low bit-width quantization problem. For
2-bit (ternary) CNNs, the quantization of N weights can be done by an exact formula
in O(N log N) complexity. When the bit-width is 3 and above, we further propose a
semi-analytical thresholding scheme with a single free parameter for quantization that is
computationally inexpensive. The free parameter is further determined by network retraining
and object detection tests. The LBW-Net has several desirable advantages over
full-precision CNNs, including considerable memory savings, energy efficiency, and faster
deployment. Our experiments on PASCAL VOC dataset show that compared with its
32-bit floating-point counterpart, the performance of the 6-bit LBW-Net is nearly lossless
in the object detection tasks, and can even do better in real world visual scenes, while
empirically enjoying more than 4× faster deployment.
more »
« less
Neural Network Training with Approximate Logarithmic Computations
The high computational complexity associated with training deep neural networks limits online and real-time training on edge devices. This paper proposed an end-to-end training and inference scheme that eliminates multiplications by approximate operations in the log-domain which has the potential to significantly reduce implementation complexity. We implement the entire training procedure in the log-domain, with fixed-point data representations. This training procedure is inspired by hardware-friendly approximations of log-domain addition which are based on look-up tables and bit-shifts. We show that our 16-bit log-based training can achieve classification accuracy within approximately 1% of the equivalent floating-point baselines for a number of commonly used datasets.
more »
« less
- Award ID(s):
- 1763747
- PAR ID:
- 10163070
- Date Published:
- Journal Name:
- ICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)
- Page Range / eLocation ID:
- 3122 to 3126
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
null (Ed.)Portfolio-based algorithm selection has seen tremendous practical success over the past two decades. This algorithm configuration procedure works by first selecting a portfolio of diverse algorithm parameter settings, and then, on a given problem instance, using an algorithm selector to choose a parameter setting from the portfolio with strong predicted performance. Oftentimes, both the portfolio and the algorithm selector are chosen using a training set of typical problem instances from the application domain at hand. In this paper, we provide the first provable guarantees for portfolio-based algorithm selection. We analyze how large the training set should be to ensure that the resulting algorithm selector's average performance over the training set is close to its future (expected) performance. This involves analyzing three key reasons why these two quantities may diverge: 1) the learning-theoretic complexity of the algorithm selector, 2) the size of the portfolio, and 3) the learning-theoretic complexity of the algorithm's performance as a function of its parameters. We introduce an end-to-end learning-theoretic analysis of the portfolio construction and algorithm selection together. We prove that if the portfolio is large, overfitting is inevitable, even with an extremely simple algorithm selector. With experiments, we illustrate a tradeoff exposed by our theoretical analysis: as we increase the portfolio size, we can hope to include a well-suited parameter setting for every possible problem instance, but it becomes impossible to avoid overfitting.more » « less
-
null (Ed.)Speech enhancement techniques that use a generative adversarial network (GAN) can effectively suppress noise while allowing models to be trained end-to-end. However, such techniques directly operate on time-domain waveforms, which are often highly-dimensional and require extensive computation. This paper proposes a novel GAN-based speech enhancement method, referred to as S-ForkGAN, that operates on log-power spectra rather than on time-domain speech waveforms, and uses a forked GAN structure to extract both speech and noise information. By operating on log-power spectra, one can seamlessly include conventional spectral subtraction techniques, and the parameter space typically has a lower dimension. The performance of S-ForkGAN is assessed for automatic speech recognition (ASR) using the TIMIT data set and a wide range of noise conditions. It is shown that S-ForkGAN outperforms existing GAN-based techniques and that it has a lower complexity.more » « less
-
Emerging resistive random-access memory (ReRAM) has recently been intensively investigated to accelerate the processing of deep neural networks (DNNs). Due to the in-situ computation capability, analog ReRAM crossbars yield significant throughput improvement and energy reduction compared to traditional digital methods. However, the power hungry analog-to-digital converters (ADCs) prevent the practical deployment of ReRAM-based DNN accelerators on end devices with limited chip area and power budget. We observe that due to the limited bitdensity of ReRAM cells, DNN weights are bit sliced and correspondingly stored on multiple ReRAM bitlines. The accumulated current on bitlines resulted by weights directly dictates the overhead of ADCs. As such, bitwise weight sparsity rather than the sparsity of the full weight, is desirable for efficient ReRAM deployment. In this work, we propose bit-slice `1, the first algorithm to induce bit-slice sparsity during the training of dynamic fixed-point DNNs. Experiment results show that our approach achieves 2 sparsity improvement compared to previous algorithms. The resulting sparsity allows the ADC resolution to be reduced to 1-bit of the most significant bit-slice and down to 3-bit for the others bits, which significantly speeds up processing and reduces power and area overhead.more » « less
-
null (Ed.)We consider the communication complexity of a number of distributed optimization problems. We start with the problem of solving a linear system. Suppose there is a coordinator together with s servers P1, …, Ps, the i-th of which holds a subset A(i) x = b(i) of ni constraints of a linear system in d variables, and the coordinator would like to output an x ϵ ℝd for which A(i) x = b(i) for i = 1, …, s. We assume each coefficient of each constraint is specified using L bits. We first resolve the randomized and deterministic communication complexity in the point-to-point model of communication, showing it is (d2 L + sd) and (sd2L), respectively. We obtain similar results for the blackboard communication model. As a result of independent interest, we show the probability a random matrix with integer entries in {–2L, …, 2L} is invertible is 1–2−Θ(dL), whereas previously only 1 – 2−Θ(d) was known. When there is no solution to the linear system, a natural alternative is to find the solution minimizing the ℓp loss, which is the ℓp regression problem. While this problem has been studied, we give improved upper or lower bounds for every value of p ≥ 1. One takeaway message is that sampling and sketching techniques, which are commonly used in earlier work on distributed optimization, are neither optimal in the dependence on d nor on the dependence on the approximation ε, thus motivating new techniques from optimization to solve these problems. Towards this end, we consider the communication complexity of optimization tasks which generalize linear systems, such as linear, semi-definite, and convex programming. For linear programming, we first resolve the communication complexity when d is constant, showing it is (sL) in the point-to-point model. For general d and in the point-to-point model, we show an Õ(sd3L) upper bound and an (d2 L + sd) lower bound. In fact, we show if one perturbs the coefficients randomly by numbers as small as 2−Θ(L), then the upper bound is Õ(sd2L) + poly(dL), and so this bound holds for almost all linear programs. Our study motivates understanding the bit complexity of linear programming, which is related to the running time in the unit cost RAM model with words of O(log(nd)) bits, and we give the fastest known algorithms for linear programming in this model. Read More: https://epubs.siam.org/doi/10.1137/1.9781611975994.106more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- https://doi.org/10.1109/ICASSP40776.2020.9053015
