研究者詳細 - 田向　権

INT8 Activation Ternary or Binary Weights Networks 査読有り

Fuengfusin Ninnart, Tamukoh Hakaru

人工生命とロボットに関する国際会議予稿集 ( 株式会社ALife Robotics ) 27 599 - 603 2022年01月

担当区分：最終著者記述言語：英語掲載種別：研究論文（学術雑誌）

We propose binary or ternary weights with 8-bit integer activation convolutional neural networks. This model is designed as a middle ground between 8-bit integer and 1-bit or 2-bit quantized models. We discover that conventional 1-bit or 2-bit only-weight quantization methods (i.e., BinaryConnect and Ternary weights network) can be utilized jointly with 8-bit integer activation quantization without significant fractions. Based on these two methods, we evaluate our model with a VGG16-like model and CIFAR10 dataset. Our model provides competitive results to a conventional floating-point model.

DOI： 10.5954/icarob.2022.os15-1

Emergence of In-Materio Intelligence from an Incidental Structure of a Single-Walled Carbon Nanotube–Porphyrin Polyoxometalate Random Network 査読有り国際誌

Deep Banerjee,Takumi Kotooka,Saman Azhari,Yuki Usami,Takuji Ogawa,James K. Gimzewski,Hakaru Tamukoh,Hirofumi Tanaka

Advanced Intelligent Systems 4 ( 4 ) 2022年01月

記述言語：英語掲載種別：研究論文（学術雑誌）

Hardware-based machine intelligence with the network architecture of reservoir computing (RC) is gaining interest because of its biological computational resemblance along with an easy and efficient neural network training approach. Herein, such a physical RC (in-materio RC) platform consisting of a recurrent network formed by the single-walled carbon nanotube (SWNT)–porphyrin polyoxometalate (Por–POM) complex is demonstrated. The network architecture executes the fundamental reservoir properties of nonlinearity, higher harmonic generation, and 1/fγ power law information processing ability. Based on these functionalities, an RC benchmark task of waveform generation is performed where the device achieves maximum fitting accuracy of 99.4%. Furthermore, a supervised object classification task based on a one-hot vector target is also executed using Toyota Human Support Robot tactile inputs. The successful classification of objects of different hardness is enhanced when the device output response follows the 1/fγ power law of maximized information processing.

DOI： 10.1002/aisy.202100145

An Implementation Method Using Cut-Off Bits for Restricted Boltzmann Machines Without Random Number Generators 査読有り

Hori S., Tamukoh H.

IEEE Access 10 42791 - 42801 2022年01月

担当区分：最終著者記述言語：英語掲載種別：研究論文（学術雑誌）

This study proposes an implementation method of a hardware-oriented restricted Boltzmann machine (RBM) without random number generators (RNGs) that employ cut-off bits, which are obtained from fixed-point binary arithmetic operations on digital hardware, such as field-programmable gate arrays (FPGAs), instead of random numbers. Most FPGA circuits employ fixed-point binary arithmetic operations to improve hardware resource efficiency. Therefore, the proposed method applies the unique feature of the operation, which is bit width extension and cut-off bits. Stochastic neural networks, including RBMs, employ sampling processes based on a probability distribution associated with the network, and the processes require many random numbers. However, implementing RNGs in hardware is costly because it requires considerable hardware resources. The proposed method can mitigate this requirement. To validate the proposed method, we implement an RBM with the proposed method on the software, emulate fixed-point binary arithmetic operations, and train the RBM using the MNIST and Fashion MNIST datasets. Furthermore, we apply the chi-square goodness-of-fit test to evaluate the uniformity of the cut-off bits. Additionally, we compare hardware resource requirements and power consumption for the proposed method and some major RNGs, a linear feedback shift register (LFSR), and a xorshift. Experimental results showed that it was possible to use the cut-off bits for training the RBM using the datasets and clarified the properties of the cut-off bits using statistical analyses. Moreover, hardware implementation of the proposed method involved the lowest hardware resource requirements and power consumption among the RNGs compared in this study.

DOI： 10.1109/ACCESS.2022.3168026

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85129145824&origin=inward

Solution of World Robot Challenge 2020 Partner Robot Challenge (Real Space) 査読有り

Ono T., Kanaoka D., Shiba T., Tokuno S., Yano Y., Mizutani A., Matsumoto I., Amano H., Tamukoh H.

Advanced Robotics 36 ( 17-18 ) 870 - 889 2022年01月

担当区分：最終著者記述言語：英語掲載種別：研究論文（学術雑誌）

The World Robot Challenge is an international competition for the social implementation of robots. Among them, the Partner Robot Challenge (Real Space) is a category that focuses on domestic service robots, competing to achieve simple tasks such as tidying a room, avoiding small obstacles, grasping a specified object from the shelf, and delivering an object to a waving person. In this category, we focused on the theme of ‘Keep Moving’ and worked on researching and developing technologies for object recognition, grasping, and other tasks. For object recognition, we propose an automatic dataset generator using a physics simulator and generate about 490,000 images in two hours. We trained it with an instance segmentation model, achieving an accuracy of about 79.3% in all games. For object grasping, we installed a three-dimensional sensor on the robot hand, and achieved successful grasping with high accuracy of about 79.6% (88.9% excluding failures due to hardware problems) in all games. In addition, we propose a motion synthesis method that simultaneously performs movement and posture transition to achieve high-speed motion. When navigation is executed, the robot changes to the navigation posture while aiming for the goal. Before reaching the goal, it changes the posture for the following action. As a result, we realized a speedup of about 1.32 times in the pick-and-place task. In the tidy-up task, the robot could clear an average of 14 objects and a maximum of 16 objects in 15 minutes. In the obstacle avoidance task, the robot succeeded five times out of six games, and in the delivery task, it succeeded four times out of six games. Throughout all the tasks, we scored 800 points twice, the highest score in the world, and won the championship. This research is aimed at the practical application of domestic service robots. We believe that the results presented in this paper demonstrate the practicality of domestic service robots in a well-developed environment. The video of the finals is available at https://youtu.be/ElUb8bfSC34?t=5511 (right side). The our source code using the simulator is available at https://github.com/Hibikino-Musashi-Home/hma_wrs_sim_ws.

DOI： 10.1080/01691864.2022.2115315

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85137831520&origin=inward

Reservoir-based convolution 査読有り

Tanaka Yuichiro, Tamukoh Hakaru

Nonlinear Theory and Its Applications, IEICE ( 一般社団法人電子情報通信学会 ) 13 ( 2 ) 397 - 402 2022年01月

担当区分：最終著者記述言語：英語掲載種別：研究論文（学術雑誌）

<p>Reservoir computing (RC) has attracted attention and has been used in many applications because of its low training cost. Multiple studies using RC for image recognition have been proposed, and some have achieved accuracy rates of greater than 99% on the MNIST dataset. For the Fashion-MNIST and CIFAR-10 datasets, however, they have not yet achieved high accuracy. This study proposes a novel convolutional neural network based on RC that can be optimized by ridge regression rather than back-propagation. The reservoir-based network has multiple reservoirs with various leak rates to extract features with various spatial frequencies from the inputs. The experimental results show that the performance of the proposed model achieves higher accuracy rates in the mentioned datasets compared with those of other reservoir-based image recognition approaches.</p>

DOI： 10.1587/nolta.13.397

その他リンク： https://kyutech.repo.nii.ac.jp/records/7914

Applying Center Loss to Multidimensional Feature Space in Deep Neural Networks for Open-set Recognition 査読有り

Kanaoka D., Tanaka Y., Tamukoh H.

Proceedings of the International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications 5 359 - 365 2022年01月

担当区分：最終著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

With the advent of deep learning, significant improvements in image recognition performance have been achieved. In image recognition, it is generally assumed that all the test data are composed of known classes. This approach is termed as closed-set recognition. In closed-set recognition, when an untrained, unknown class is input, it is recognized as one of the trained classes. The method whereby an unknown image is recognized as unknown when it is input is termed as open-set recognition. Although several open-set recognition methods have been proposed, none of these previous methods excel in terms of all three evaluation items: learning cost, recognition performance, and scalability from closed-set recognition models. To address this, we propose an open-set recognition method using the distance between features in the multidimensional feature space of neural networks. By applying center loss to the feature space, we aim to maintain the classification accuracy of closed-set recognition and improve the unknown detection performance. In our experiments, we achieved state-of-the-art performance on the MNIST, SVHN, and CIFAR-10 datasets. In addition, the proposed approach shows excellent performance in terms of the three evaluation items.

DOI： 10.5220/0010816600003124

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85143999490&origin=inward

Hardware-oriented Algorithm for Human Detection using GMM-MRCoHOG Features 査読有り

Takemoto R., Nagamine Y., Yoshihiro K., Shibata M., Yamada H., Tanaka Y., Enokida S., Tamukoh H.

Proceedings of the International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications 4 749 - 757 2022年01月

担当区分：最終著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

In this research, we focus on Gaussian mixture model-multiresolution co-occurrence histograms of oriented gradients (GMM-MRCoHOG) features using luminance gradients in images and propose a hardware-oriented algorithm of GMM-MRCoHOG to implement it on a field programmable gate array (FPGA). The proposed method simplifies the calculation of luminance gradients, which is a high-cost operation in the conventional algorithm, by using lookup tables to reduce the circuit size. We also designed a human-detection digital architecture of the proposed algorithm for FPGA implementation using high-level synthesis. The verification results showed that the processing speed of the proposed architecture was approximately 123 times faster than that of the FPGA implementation of VGG-16.

DOI： 10.5220/0010848100003124

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85182929508&origin=inward

A Control and Data Transfer Platform for FPGA Applications 査読有り

Hori S., Tamukoh H.

Journal of Robotics, Networking and Artificial Life ( 株式会社 ALife Robotics ) 9 ( 1 ) 37 - 42 2022年01月

担当区分：最終著者記述言語：英語掲載種別：研究論文（学術雑誌）

Deep learning technology has made remarkable progress in recent years and has been adopted for a variety of applications such as smartphones and cloud servers. These systems employ dedicated processors to reduce power consumption and process massive amounts of data. In this paper, we propose a field-programmable gate array (FPGA) infrastructure for easy verification of user logic. The infrastructure makes it easy to communicate and control a host PC and user logic. We implemented two example logics, which were simple image processing and a restricted Boltzmann machine, into the infrastructure to confirm these features.

DOI： 10.57417/jrnal.9.1_37

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85178265499&origin=inward

Desgin and Implementation of ROS2-based Autonomous Tiny Robot Car with Integration of Multiple ROS2 FPGA Nodes 査読有り

Mori H., Amano H., Mizutani A., Okazaki E., Konno Y., Sada K., Ono T., Yoshimoto Y., Tamukoh H., Ohkawa T., Sugaya M.

FPT 2022 - 21st International Conference on Field-Programmable Technology, Proceedings 2022年01月

記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

This paper introduces an autonomous tiny robot car equipped with a camera-based lane detection function and a traffic signal/obstacle, pedestrian recognition function. Each function is integrated by Robot Operating System 2 (ROS2), a middleware for robot system development. Autonomous driving without the need for a driver requires not only lane-following driving but also traffic signal recognition and obstacle recognition. These functions are implemented on FPGA, and we evaluated them. According to these results, the execution time of traffic signal recognition by FPGA was 1.2 to 3.4 times faster than CPU execution. YOLOv4 is used for obstacle recognition, which improved mAP by 3.79 points compared to YOLO v3-Tiny.

DOI： 10.1109/ICFPT56656.2022.9974433

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85145559746&origin=inward

INT8 Activation Ternary or Binary Weights Networks: Unifying Between INT8 and Lower-bit Width Quantization 査読有り

Fuengfusin N., Tamukoh H.

Journal of Robotics, Networking and Artificial Life ( 株式会社 ALife Robotics ) 9 ( 2 ) 171 - 176 2022年01月

担当区分：最終著者記述言語：英語掲載種別：研究論文（学術雑誌）

This paper proposes ternary or binary weights with 8-bit integer activation convolutional neural networks. Our proposed model serves as the middle ground between 8-bit integer and lower than 8-bit precision quantized models. Our empirical experiments established that the conventional 1-bit or 2-bit only-weight quantization methods (i.e., BinaryConnect and ternary weights network) can be used jointly with the 8-bit integer activation quantization. We evaluate our model with the VGG16-like model to operate with the CIFAR10 and CIFAR100 datasets. Our models show competitive results to the general 32-bit floating point model.

DOI： 10.57417/jrnal.9.2_171

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85176926162&origin=inward

INT8 Activation Ternary or Binary Weights Networks 査読有り

Fuengfusin N., Tamukoh H.

Proceedings of International Conference on Artificial Life and Robotics 599 - 603 2022年01月

担当区分：最終著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）