S. Ambrogio, P. Narayanan, H. Tsai, C. Mackin, K. Spoon, An Chen, A. Fasoli, A. Friz, G. Burr
{"title":"Accelerating Deep Neural Networks with Analog Memory Devices","authors":"S. Ambrogio, P. Narayanan, H. Tsai, C. Mackin, K. Spoon, An Chen, A. Fasoli, A. Friz, G. Burr","doi":"10.1109/AICAS48895.2020.9073978","DOIUrl":"https://doi.org/10.1109/AICAS48895.2020.9073978","url":null,"abstract":"Acceleration of training and inference of Deep Neural Networks (DNNs) with non-volatile memory (NVM) arrays, such as Phase-Change Memory (PCM), shows promising advantages in terms of energy efficiency and speed with respect to digital implementations using CPUs and GPUs. By leveraging a combination of PCM devices and CMOS circuits, high training accuracy can be achieved, leading to software-equivalent results on small and medium datasets. In addition, weights encoded with multiple PCM devices can lead to high speed and low-power inference, as shown here for Long-Short Term Memory (LSTM) networks.","PeriodicalId":145034,"journal":{"name":"2020 2nd IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS)","volume":"19 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116245001","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Distributed Clique-Based Neural Networks for Data Fusion at the Edge","authors":"Benoît Larras, A. Frappé","doi":"10.1109/AICAS48895.2020.9073801","DOIUrl":"https://doi.org/10.1109/AICAS48895.2020.9073801","url":null,"abstract":"Distributed smart sensors are more and more used in applications such as biomedical or domestic monitoring. However, each sensor broadcasts data wirelessly to the others or to an aggregator, which leads to energy-hungry sensors not ensuring data privacy. To tackle both challenges, this work proposes to distribute a part of a clique-based neural network in each sensor. This scheme allows standardizing data at the sensor level, ensuring privacy if the data is intercepted. Besides, a lower number of bits is transmitted, thus limiting the communication overhead. The circuit implementation is possible with the use of single-cluster iterative clique-based circuits. To this end, a hardware circuit has been fabricated and performs a classification using 115fJ per synaptic event per neuron in 83ns.","PeriodicalId":145034,"journal":{"name":"2020 2nd IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS)","volume":"176 7","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"120919504","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"A Real-Time Robot-Arm Surgical Guiding System Development by Image-Tracking","authors":"Ching-Hwa Cheng","doi":"10.1109/AICAS48895.2020.9073805","DOIUrl":"https://doi.org/10.1109/AICAS48895.2020.9073805","url":null,"abstract":"The endoscope is widely used for various diagnoses and treatments in Minimally Invasive Surgery (MIS), such as hysteroscopy, laparoscopy, and colonoscopy. However, the limited field of image of the endoscope is often the most problematic issue faced by surgeons and medical students, especially for those inexperienced physicians, which leads to difficulty during surgical operations. To reduce the difficulties of MIS with respect to endoscope function, the proposed identifying and locating techniques provide the angle and distance from the surgical instruments to the lesion. The in-time guiding information provides global positioning information by tracking the lesion position during surgery. The jointed with a robot-arm system can help an inexperienced surgeon with stable assistance for the long-time surgical operation. The whole system has been successfully validated by surgeons.","PeriodicalId":145034,"journal":{"name":"2020 2nd IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS)","volume":"68 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121457569","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Alpha Renner, Matthew Evanusa, G. Orchard, Yulia Sandamirskaya
{"title":"Event-Based Attention and Tracking on Neuromorphic Hardware","authors":"Alpha Renner, Matthew Evanusa, G. Orchard, Yulia Sandamirskaya","doi":"10.1109/AICAS48895.2020.9073789","DOIUrl":"https://doi.org/10.1109/AICAS48895.2020.9073789","url":null,"abstract":"We present a fully event-driven vision and processing system for selective attention and tracking implemented on Intel’s neuromorphic research chip, Loihi, directly interfaced with an event-based Dynamic Vision Sensor, DAVIS. The attention mechanism is realized as a recurrent spiking neural network (SNN) that forms sustained activation-bump attractors. The network dynamics support object tracking when distractors are present and when the object slows down or stops.","PeriodicalId":145034,"journal":{"name":"2020 2nd IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS)","volume":"10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132166851","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Jiun-Wei Yang, Peng-Ren Lai, Pin-Chieh Fu, Jia-Shung Wang
{"title":"Tradeoff between Parallel Efficiency and Coding Efficiency of HEVC: A Load Balancing Solution with Convolutional Neural Networks","authors":"Jiun-Wei Yang, Peng-Ren Lai, Pin-Chieh Fu, Jia-Shung Wang","doi":"10.1109/AICAS48895.2020.9073933","DOIUrl":"https://doi.org/10.1109/AICAS48895.2020.9073933","url":null,"abstract":"The concept of tiles was introduced in HEVC as a parallelization strategy. Tiles in HEVC enable the video to be split into independent rectangular regions, thus can be processed in parallel. However, as more threads are devoted to the execution of a parallelized program, the load imbalance among tiles would make threads’ idle time to be expected to increase, thus harm the aggregated coding time and speedup performance as well. To expose as much parallelism of HEVC encoding as possible, in this paper, several skills including CNN are conducted. Our design is based on the tradeoff between parallel efficiency and coding efficiency. The goal of CNN filtering are two folds. Firstly, the CNN model gives the prediction of the quadtree structure for each CTU, avoiding the time-consuming CU partition in HEVC if the prediction is reliable. In addition, these CNN outputs (probabilities) can be employed as a filter to decide whether a CTU will be processed by a low-complexity CNN partition mode or the original HEVC mode. The experimental results show that the proposed method could perform well in terms of speedup and aggregated coding time with negligible (minor) coding efficiency loss.","PeriodicalId":145034,"journal":{"name":"2020 2nd IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS)","volume":"109 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114500825","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Mauro Mangia, Alex Marchioni, Luciano Prono, Fabio Pareschi, R. Rovatti, G. Setti
{"title":"Low-power ECG acquisition by Compressed Sensing with Deep Neural Oracles","authors":"Mauro Mangia, Alex Marchioni, Luciano Prono, Fabio Pareschi, R. Rovatti, G. Setti","doi":"10.1109/AICAS48895.2020.9073945","DOIUrl":"https://doi.org/10.1109/AICAS48895.2020.9073945","url":null,"abstract":"The recovery of sparse signals from their linear mapping on a lower-dimensional space is traditionally performed by finding the sparsest solution compatible with such solutions. This task can be partitioned in two phases: support estimation and coefficient estimation. We propose to perform the former with a deep neural network jointly trained with the encoder that divines a support that is used in the latter phase to estimate the coefficients by pseudo-inversion. Numerical evidence demonstrates that the proposed encoder-decoder architecture outperforms state-of-the-art Compressed Sensing (CS) approaches in the recovery of synthetic ECG signals for a compression ratio higher than 2.5. Further tests on real ECG prove the applicability in real-world scenarios.","PeriodicalId":145034,"journal":{"name":"2020 2nd IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS)","volume":"56 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114723432","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Improving the Robustness of Threshold-Based Single Hidden Layer Neural Networks via Regularization","authors":"E. Ragusa, C. Gianoglio, R. Zunino, P. Gastaldo","doi":"10.1109/AICAS48895.2020.9073976","DOIUrl":"https://doi.org/10.1109/AICAS48895.2020.9073976","url":null,"abstract":"Edge computing can take full advantage of data-driven models only if the eventual inference function can be deployed on low-power, resource-constrained digital devices. In this regard, single layer feedforward neural networks (SLFNs) based on the threshold activation function represent a suitable option, as they can suitably balance generalization performances and computational costs. Their robustness to perturbations at the node level, though, is an issue: a small perturbation affecting the input to the activation might result in a sign inversion at the output of the neuron. This in turn may severely affect the accuracy of the inference function when implemented on hardware. This paper shows that the robustness of this class of SLFNs can be improved by introducing in the cost function a regularization term specifically designed to limit the impacts of perturbations at the node level. The novel cost function indeed admits a closed-form solution. Experimental validation involved six real world benchmarks.","PeriodicalId":145034,"journal":{"name":"2020 2nd IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS)","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123179986","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Retinal Slip Estimation and Object Tracking with an Active Event Camera","authors":"Qingpeng Zhu, J. Triesch, Bertram E. Shi","doi":"10.1109/AICAS48895.2020.9073922","DOIUrl":"https://doi.org/10.1109/AICAS48895.2020.9073922","url":null,"abstract":"This paper presents a retinal slip estimation algorithm and a novel tracking strategy for active event-based cameras. Event-based sensors have independent, asynchronous pixels that report local luminance change. Unlike conventional image sensors, the event cameras have a high dynamic range and high temporal resolution. To preserve these advantages, we propose an algorithm that enables estimates of the normal component of the optical flow to be updated as spikes come in, eschewing the accumulation of events within a fixed temporal window. We also propose a method for integrating these normal flow estimates to estimate the retinal slip between the camera and object. We demonstrate the advantages of this asynchronous estimator of the retinal slip using an active event camera simulator (AESIM), which we developed to enable simulation of the active control of an event sensor in a dynamic environment.","PeriodicalId":145034,"journal":{"name":"2020 2nd IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS)","volume":"120 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128032764","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Online Extreme Learning Machine Design for the Application of Federated Learning","authors":"Yi-Ta Chen, Yu-Chuan Chuang, A. Wu","doi":"10.1109/AICAS48895.2020.9073802","DOIUrl":"https://doi.org/10.1109/AICAS48895.2020.9073802","url":null,"abstract":"In this paper, we propose a federated extreme learning machine system (Fed-ELMS) to meet the demand for federated learning scenarios. In the scenario of federated learning, data is kept on edge devices to preserve the privacy of data, while metadata, such as model parameters, are exchanged between a centralized cloud server and edge devices. Despite non-independent and identically distributed (non-IID) and imbalanced data across edge devices, we show that Fed-ELMS can still achieve comparable performance with only 3.3% accuracy loss compared to a centralized ELM trained with IID and balanced data. (a) Moreover, by quantizing input weights and biases, parameters of a model and transmission power consumption between cloud and edge are dramatically reduced. Compared With conventional neural networks (NNs) with the same transmission cost, the proposed Fed-ELMS outperforms FederatedAveraging NN (Fed- NN) by 2.3% accuracy and the fine-tuning process is 7%-33% less time-consuming. Therefore, the proposed Fed-ELMS is a promising system for edge devices to support the future trend of federated learning.","PeriodicalId":145034,"journal":{"name":"2020 2nd IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131113666","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Xinxin Wang, Qiwen Wang, F. Meng, Seung Hwan Lee, Wei D. Lu
{"title":"Deep Neural Network Mapping and Performance Analysis on Tiled RRAM Architecture","authors":"Xinxin Wang, Qiwen Wang, F. Meng, Seung Hwan Lee, Wei D. Lu","doi":"10.1109/AICAS48895.2020.9073942","DOIUrl":"https://doi.org/10.1109/AICAS48895.2020.9073942","url":null,"abstract":"Representative deep neural networks (DNNs) have been successfully mapped on an RRAM-based tiled in-memory computing (IMC) architecture. Effects of finite array size and quantized partial products (PPs) due to ADC precision constraints have been analyzed. Methods were developed to solve these challenges and preserve DNN accuracies and IMC performance gains in the tiled architecture. Popular models VGG-16, MobileNet, and RNN/LSTM have been successfully implemented and tested on ImageNet dataset and text classification tasks, respectively.","PeriodicalId":145034,"journal":{"name":"2020 2nd IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS)","volume":"6 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132573986","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}