In the annual meeting of IEEE Kansai Section held on February 22nd, 2019, Yuki Tanaka received the IEEE Kansai Section Student Paper Award for his following publication.
Lab tour was held on 1pm-3pm Feb. 21, 2019. Please feel free to visit us at any time or send an e-mail for an appointment, if you missed the chance.
The following paper has been accepted for presentation in DAC2019（ACM/IEEE Design Automation Conference 2019） (acceptance rate 24.8%=202/815). Conguratulations!!
A Ph.D student, Song gave an oral presentation in Asia and South Pacific Design Automation Conference (ASP-DAC) 2019 held at the National Musium of Science and Art during Jan. 21-24, 2019. (The presentation was on Jan. 25th.)
His presentation is about a secure filter, specically, a secure version of the naive Bayesian filter (NBF) is proposed utilizing partially homomorphic encryption (PHE) scheme. SNBF can be implemented with only the additive homomorphism from the Paillier system, and we derive new techniques to reduce the computational cost of PHE-based SNBF. In the experiment, we implemented SNBF both in software and hardware. Compared to the best existing PHE scheme, we achieved 1,200x and 398,840x runtime reduction for CPU and ASIC implementations, repsectively, with additional 1,919x power reduction on the designated hardware multiplier. Our hardware implementation is able to classify an average-length email in 0.5s, making it one of the most practical NBF schemes to date.
Sorry, this entry is only available in 日本語.
Prof. Yiyu Shi of Notre Dame University presented his recent work on statistical neural network. He is currently with our lab for a JSPS fellowship program.
SCNN: A General Distribution based Statistical Convolutional Neural Network with Application to Video Object Detection
Abstract: Various convolutional neural networks (CNNs) were developed recently that achieved accuracy comparable with that of human beings in computer vision tasks such as image recognition, object detection and tracking, etc. Most of these networks, however, process one single frame of image at a time, and may not fully utilize the temporal and contextual correlation typically present in multiple channels of the same image or adjacent frames from a video, thus limiting the achievable throughput. This limitation stems from the fact that existing CNNs operate on deterministic numbers. In this talk, I will present a novel statistical convolutional neural network (SCNN), which extends existing CNN architectures but operates directly on correlated distributions rather than deterministic numbers. By introducing a parameterized canonical model to model correlated data and defining corresponding operations as required for CNN training and inference, I will show that SCNN can process multiple frames of correlated images effectively, hence achieving significant speedup over existing CNN models. I will use a CNN based video object detection as an example to illustrate the usefulness of the proposed SCNN as a general network model. Experimental results show that even a non-optimized implementation of SCNN can still achieve 178% speedup over existing CNNs with slight accuracy degradation.
Welcome party as well as end of year party was held in Tentra restaurant on Dec. 17, 2018. Prof. Yiyu Shi (Notre Dame university) and special auditor student, Andregg Raphael (ETH Zurich) are currently visiting our lab. Over shabu-shabu dinner, we really enjoyed stimulating conversations. Cheers!
Two papers have been accepted for presentation in ICMTS 2019 (International Conference on Microelectronic Test Structures).
The following paper has been accepted for publication in Japanese Journal of Applied Physics (JJAP). This is a joint work with National Institute of Advanced Industrial Science.
