Design of a 2-Bit Neural Network Quantizer for Laplacian Source

Achieving real-time inference is one of the major issues in contemporary neural network applications, as complex algorithms are frequently being deployed to mobile devices that have constrained storage and computing power. Moving from a full-precision neural network model to a lower representation by applying quantization techniques is a popular approach to facilitate this issue. Here, we analyze in detail and design a 2-bit uniform quantization model for Laplacian source due to its significance in terms of implementation simplicity, which further leads to a shorter processing time and faster inference. The results show that it is possible to achieve high classification accuracy (more than 96% in the case of MLP and more than 98% in the case of CNN) by implementing the proposed model, which is competitive to the performance of the other quantization solutions with almost optimal precision.

Implementation of generative adversarial networks in HPCC systems using GNN bundle

<span id="docs-internal-guid-8df40bfb-7fff-fcb8-b52f-54f39570d649"><span>HPCC systems, an open source cluster computing platform for big data analytics consists of generalized neural network bundle with a wide variety of features which can be used for various neural network applications. To enhance the functionality of the bundle, this paper proposes the design and development of generative adversarial networks (GANs) on HPCC systems platform using ECL, a declarative language on which HPCC systems works. GANs have been developed on the HPCC platform by defining the generator and discriminator models separately, and training them by batches in the same epoch. In order to make sure that they train as adversaries, a certain weights transfer methodology was implemented. MNIST dataset which has been used to test the proposed approach has provided satisfactory results. The results obtained were unique images very similar to the MNIST dataset, as it were expected.</span></span>