This paper proposes ReBNet, an end-to-end framework for training reconfigurable binary neural networks on software and developing efficient accelerators for execution on FPGA. Binary neural networks offer an intriguing opportunity for deploying large-scale deep learning models on resource-constrained devices. Binarization reduces the memory footprint and replaces the power-hungry matrix-multiplication with light-weight XnorPopcount operations. However, binary networks suffer from a degraded accuracy compared to their fixed-point counterparts. We show that the state-of-the-art methods for optimizing binary networks accuracy, significantly increase the implementation cost and complexity. To compensate for the degraded accuracy while adhering to the simplicity of binary networks, we devise the first reconfigurable scheme that can adjust the classification accuracy based on the application. Our proposition improves the classification accuracy by representing features with multiple levels of residual binarization. Unlike previous methods, our approach does not exacerbate the area cost of the hardware accelerator. Instead, it provides a tradeoff between throughput and accuracy while the area overhead of multi-level binarization is negligible.
We propose a stacked 1D convolutional neural network (S1DCNN) for end-to-end small footprint voice trigger detection in a streaming scenario. Voice trigger detection is an important speech application, with which users can activate their devices by simply saying a keyword or phrase. Due to privacy and latency reasons, a voice trigger detection system should run on an always-on processor on device. Therefore, having small memory and compute cost is crucial for a voice trigger detection system. Recently, singular value decomposition filters (SVDFs) has been used for end-to-end voice trigger detection. The SVDFs approximate a fully-connected layer with a low rank approximation, which reduces the number of model parameters. In this work, we propose S1DCNN as an alternative approach for end-to-end small-footprint voice trigger detection. An S1DCNN layer consists of a 1D convolution layer followed by a depth-wise 1D convolution layer. We show that the SVDF can be expressed as a special case of the S1DCNN layer. Experimental results show that the S1DCNN achieve 19.0% relative false reject ratio (FRR) reduction with a similar model size and a similar time delay compared to the SVDF. By using longer time delays, the S1DCNN further improve the FRR up to 12.2% relative.
Binary Decision Diagram (BDD) is a modern data structure proved to be compact in representation and efficient in manipulation of Boolean formulas. Using Binary decision diagram in network reliability analysis has already been investigated by some researchers. In this paper we show how an exact algorithm for network reliability can be improved and implemented efficiently using CUDD -Colorado University Decision Diagram.
This research work is concerned with the predictability of ensemble and singular tree-based machine learning algorithms during the recession and prosperity of the two companies listed in the Tehran Stock Exchange in the context of big data. In this regard, the main issue is that economic managers and the academic community require predicting models with more accuracy and reduced execution time; moreover, the prediction of the companies recession in the stock market is highly significant. Machine learning algorithms must be able to appropriately predict the stock return sign during the market downturn and boom days. Addressing the stated challenge will upgrade the quality of stock purchases and, subsequently, will increase profitability. In this article, the proposed solution relies on the utilization of tree-based machine learning algorithms in the context of big data. The proposed solution exploits the decision tree algorithm, which is a traditional and singular tree-based learning algorithm. Furthermore, two modern and ensemble tree-based learning algorithms, random forest and gradient boosted tree, has been utilized for predicting the stock return sign during recession and prosperity. The mentioned cases were implemented by applying the machine learning tools in python programming language and PYSPARK library that is used explicitly for the big data context. The utilized research data of the current study are the shares information of two companies of the Tehran Stock Exchange. The obtained results reveal that the applied ensemble learning algorithms have performed better than the singular learning algorithms. Additionally, adding 23 technical features to the initial data and subsequent applying of the PCA feature reduction method have demonstrated the best performance among other modes. In the meantime, it has been concluded that the initial data do not possess the proper resolution or generalizability, either during prosperity or recession.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.