Masato Motomura
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Masato Motomura has a total of 104 co-authors. Here are the top co-authors by number of shared publications:
- Kazushi Kawamura
Faculty of Science and Engineering, Waseda University, Tokyo, Japan
14 shared publications - Jaehoon Yu
Tokyo Institute of Technology, Yokohama, Japan
10 shared publications - Thiem Van Chu
Institute of Science Tokyo, Tokyo, Japan
10 shared publications - Kota Ando
Faculty of Information Science and Technology, Hokkaido University
8 shared publications - Tetsuya Asai
Hokkaido University
5 shared publications - Daichi Fujiki
Institute of Science Tokyo, Tokyo, Japan
4 shared publications
Latest Publications
- Open Access
- 6 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Evolutionary Computation
- Vol. PP (1089778X) , 1
Annealed Mean Field Descent Is Highly Effective for Quadratic Unconstrained Binary Optimization
In recent years, formulating various combinatorial optimization problems as Quadratic Unconstrained Binary Optimization (QUBO) has gained significant attention as a promising approach for efficiently obtaining optimal or near-optimal solutions. While QUBO offers a general-purpose framework, existing solvers often struggle with performance variability across different problems. This paper (i) theoretically analyzes Mean Field Annealing (MFA) and its variants, which are representative QUBO solvers, and shows that their underlying self-consistent equations do not necessarily correspond to the minimization condition of the Kullback–Leibler divergence between the mean-field approximation and the exact distribution. Moreover, the paper (ii) proposes a novel method, Annealed Mean Field Descent (AMFD), which directly minimizes this divergence to address this limitation. Through extensive experiments on five benchmark combinatorial optimization problems (Maximum Cut Problem, Maximum Independent Set Problem, Traveling Salesman Problem, Quadratic Assignment Problem, and Graph Coloring Problem), we demonstrate that AMFD exhibits superior performance in many cases and reduced problem dependence compared to state-of-the-art QUBO solvers and Gurobi—a state-of-the-art versatile mathematical optimization solver not limited to QUBO.
- 6 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Evolutionary Computation
- Vol. PP (1089778X) , 1
Open AccessAnnealed Mean Field Descent Is Highly Effective for Quadratic Unconstrained Binary Optimization
In recent years, formulating various combinatorial optimization problems as Quadratic Unconstrained Binary Optimization (QUBO) has gained significant attention as a promising approach for efficiently obtaining optimal or near-optimal solutions. While QUBO offers a general-purpose framework, existing solvers often struggle with performance variability across different problems. This paper (i) theoretically analyzes Mean Field Annealing (MFA) and its variants, which are representative QUBO solvers, and shows that their underlying self-consistent equations do not necessarily correspond to the minimization condition of the Kullback–Leibler divergence between the mean-field approximation and the exact distribution. Moreover, the paper (ii) proposes a novel method, Annealed Mean Field Descent (AMFD), which directly minimizes this divergence to address this limitation. Through extensive experiments on five benchmark combinatorial optimization problems (Maximum Cut Problem, Maximum Independent Set Problem, Traveling Salesman Problem, Quadratic Assignment Problem, and Graph Coloring Problem), we demonstrate that AMFD exhibits superior performance in many cases and reduced problem dependence compared to state-of-the-art QUBO solvers and Gurobi—a state-of-the-art versatile mathematical optimization solver not limited to QUBO.
- Open Access
- 30 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 19424-19439
Memory-Efficient and Trustworthy Neural Networks via Random Seed-Based Design
Deep learning models deployed in resource-constrained and safety-critical settings must be not only compact and accurate but also trustworthy and capable of expressing predictive uncertainty. However, conventional compression techniques such as pruning and quantization often make models overconfident, while uncertainty-aware methods like Bayesian inference and ensembles ignore memory and compute budgets. In this work, we propose Random Seed-Based Design, a framework that elevates the random seed to a first-class design primitive for jointly shaping sparsity, freezing, and stochastic inference. Random seeds deterministically generate all stochastic components: Strong Lottery Ticket weights, partially frozen masks, and Monte Carlo Dropout masks, so that a deployed model can be exactly regenerated from seeds, binary masks, and a small set of interpretable hyperparameters instead of dense weight tensors. We formulate this seed-based construction as a multi-objective design problem over accuracy, calibration, uncertainty, FLOP, and memory footprint, and solve it with Bayesian optimization. This seed-based design enables us to characterize Pareto-optimal trade-offs between efficiency and trustworthiness across diverse architectures, including convolutional neural networks and graph neural networks, thereby illuminating domain-dependent relationships between computational efficiency and predictive reliability. Our method discovers models that are tens to hundreds of times smaller than their dense counterparts, reduces expected calibration error by up to an order of magnitude, and matches or improves accuracy. These results show that structured seed-controlled randomness improves both compression and trustworthiness, rather than treating them as competing objectives. The source code is available at https://github.com/itocha/SLT-TRUST.
- 30 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 19424-19439
Open AccessMemory-Efficient and Trustworthy Neural Networks via Random Seed-Based Design
Deep learning models deployed in resource-constrained and safety-critical settings must be not only compact and accurate but also trustworthy and capable of expressing predictive uncertainty. However, conventional compression techniques such as pruning and quantization often make models overconfident, while uncertainty-aware methods like Bayesian inference and ensembles ignore memory and compute budgets. In this work, we propose Random Seed-Based Design, a framework that elevates the random seed to a first-class design primitive for jointly shaping sparsity, freezing, and stochastic inference. Random seeds deterministically generate all stochastic components: Strong Lottery Ticket weights, partially frozen masks, and Monte Carlo Dropout masks, so that a deployed model can be exactly regenerated from seeds, binary masks, and a small set of interpretable hyperparameters instead of dense weight tensors. We formulate this seed-based construction as a multi-objective design problem over accuracy, calibration, uncertainty, FLOP, and memory footprint, and solve it with Bayesian optimization. This seed-based design enables us to characterize Pareto-optimal trade-offs between efficiency and trustworthiness across diverse architectures, including convolutional neural networks and graph neural networks, thereby illuminating domain-dependent relationships between computational efficiency and predictive reliability. Our method discovers models that are tens to hundreds of times smaller than their dense counterparts, reduces expected calibration error by up to an order of magnitude, and matches or improves accuracy. These results show that structured seed-controlled randomness improves both compression and trustworthiness, rather than treating them as competing objectives. The source code is available at https://github.com/itocha/SLT-TRUST.
- Open Access
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- p. 1
Bridging the Modeling Gap in Ising-Model-Based Optimization through Automatic Solution-Guided Refinement
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- p. 1
Open AccessBridging the Modeling Gap in Ising-Model-Based Optimization through Automatic Solution-Guided Refinement
- Open Access
- 20 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 1910-1924
BingoGCN: Towards Scalable and Efficient GNN Acceleration with Fine-Grained Partitioning and SLT
- 20 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 1910-1924
Open AccessBingoGCN: Towards Scalable and Efficient GNN Acceleration with Fine-Grained Partitioning and SLT
- Open Access
- 25 May 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-13
Trustworthy Deep Learning Acceleration with Customizable Design Flow Automation
- 25 May 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-13
Open AccessTrustworthy Deep Learning Acceleration with Customizable Design Flow Automation
- Open Access
- 12 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 86509-86527
WhiteDwarf: A Holistic Co-Design Approach to Ultra-Compact Neural Inference Acceleration
Recent work on the Strong Lottery Ticket Hypothesis has upgraded the role of sparsity, commonly exploited by algorithms and sparse architectures as an advantageous by-product of training overparameterized models, to the main driver of neural training, opening new co-design opportunities for efficient neural execution. WhiteDwarf is a holistic ultra-compact approach that first uses a triple model compression algorithm to reduce CNN and MLP models to under 10% of the original off-chip size and then uses a triple unstructured sparsity exploitation architecture for under 1% on-chip size. Folded, Signed, and Multi-coated supermasks are integrated into a novel scalar supermask that integrates pruning, quantization, and training. These supermasks offer straightforward compression and an opportunity for a sparse, mixed precision, multiplier-less datapath that does not require complex control. Activation precision is set to FP8 with large models in mind, while INT2–4 weights suffice for sparse and accurate supermask models. The 1K-PE array features parallel, hierarchically synchronous, non-zero gathering, and bit-decomposed computation for exploiting unstructured sparsity down to bit-level. For example, a ResNet-50 is compressed to 5.3% of the original model size while maintaining 74.7% accuracy on ImageNet. The fabricated 40-nm CMOS chip, aimed at high inference accuracy and power efficiency, achieves 12.24 TFLOPS/W at 99% weight sparsity.
- 12 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 86509-86527
Open AccessWhiteDwarf: A Holistic Co-Design Approach to Ultra-Compact Neural Inference Acceleration
Recent work on the Strong Lottery Ticket Hypothesis has upgraded the role of sparsity, commonly exploited by algorithms and sparse architectures as an advantageous by-product of training overparameterized models, to the main driver of neural training, opening new co-design opportunities for efficient neural execution. WhiteDwarf is a holistic ultra-compact approach that first uses a triple model compression algorithm to reduce CNN and MLP models to under 10% of the original off-chip size and then uses a triple unstructured sparsity exploitation architecture for under 1% on-chip size. Folded, Signed, and Multi-coated supermasks are integrated into a novel scalar supermask that integrates pruning, quantization, and training. These supermasks offer straightforward compression and an opportunity for a sparse, mixed precision, multiplier-less datapath that does not require complex control. Activation precision is set to FP8 with large models in mind, while INT2–4 weights suffice for sparse and accurate supermask models. The 1K-PE array features parallel, hierarchically synchronous, non-zero gathering, and bit-decomposed computation for exploiting unstructured sparsity down to bit-level. For example, a ResNet-50 is compressed to 5.3% of the original model size while maintaining 74.7% accuracy on ImageNet. The fabricated 40-nm CMOS chip, aimed at high inference accuracy and power efficiency, achieves 12.24 TFLOPS/W at 99% weight sparsity.
- 23 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
- Vol. 33 (7) , 1858-1871
DMSA: An Efficient Architecture for Sparse–Sparse Matrix Multiplication Based on Distribute-Merge Product Dataflow
The sparse–sparse matrix multiplication (SpMSpM) is a fundamental operation in various applications. Existing SpMSpM accelerators based on inner product (IP) and outer product (OP) suffer from low computational efficiency and high memory traffic due to inefficient index matching and merging overheads. Gustavson’s product (GP)-based accelerators mitigate some of these challenges but struggle with workload imbalance and irregular memory access patterns, limiting computational parallelism. To overcome these limitations, we propose a distribute-merge product (DMP), a novel SpMSpM dataflow that evenly distributes workloads across multiple computation streams and merges partial results efficiently. We design and implement DMP-based SpMSpM architecture (DMSA), incorporating four key techniques to fully exploit the parallelism of DMP and efficiently handle irregular memory accesses. Implemented on a Xilinx ZCU106 FPGA, DMSA achieves speedups of up to $3.38\times $ and $1.73\times $ over two state-of-the-art FPGA-based SpMSpM accelerators while maintaining comparable hardware resource usage. In addition, compared to CPU and GPU implementations on an NVIDIA Jetson AGX Xavier, DMSA is $4.96\times $ and $1.53\times $ faster while achieving $6.67\times $ and $2.33\times $ better energy efficiency, respectively.
- 23 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
- Vol. 33 (7) , 1858-1871
DMSA: An Efficient Architecture for Sparse–Sparse Matrix Multiplication Based on Distribute-Merge Product Dataflow
The sparse–sparse matrix multiplication (SpMSpM) is a fundamental operation in various applications. Existing SpMSpM accelerators based on inner product (IP) and outer product (OP) suffer from low computational efficiency and high memory traffic due to inefficient index matching and merging overheads. Gustavson’s product (GP)-based accelerators mitigate some of these challenges but struggle with workload imbalance and irregular memory access patterns, limiting computational parallelism. To overcome these limitations, we propose a distribute-merge product (DMP), a novel SpMSpM dataflow that evenly distributes workloads across multiple computation streams and merges partial results efficiently. We design and implement DMP-based SpMSpM architecture (DMSA), incorporating four key techniques to fully exploit the parallelism of DMP and efficiently handle irregular memory accesses. Implemented on a Xilinx ZCU106 FPGA, DMSA achieves speedups of up to $3.38\times $ and $1.73\times $ over two state-of-the-art FPGA-based SpMSpM accelerators while maintaining comparable hardware resource usage. In addition, compared to CPU and GPU implementations on an NVIDIA Jetson AGX Xavier, DMSA is $4.96\times $ and $1.53\times $ faster while achieving $6.67\times $ and $2.33\times $ better energy efficiency, respectively.
- 14 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Micro
- Vol. 45 (2) , 122-124
Derek Chiou
Derek Chiou passed away on 26 December 2024. In his multidecade career as a computer architect, Derek was exceptionally productive and influential, impacting the field from concurrent prominent positions in academia and industry. At the time of his passing, Derek was a professor in the Chandra Family Department of Electrical and Computer Engineering at the University of Texas at Austin (UT Austin) and held the Temple Foundation Endowed Teaching Fellowship in Engineering. Derek was also a partner hardware architect of Microsoft Azure. Despite these demanding roles, Derek was always generous with his time to those around him. More than his technical accomplishments, all those who have crossed paths with Derek will remember him first as an unreservedly caring and giving colleague, collaborator, mentor, teacher, and friend.
- 14 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Micro
- Vol. 45 (2) , 122-124
Derek Chiou
Derek Chiou passed away on 26 December 2024. In his multidecade career as a computer architect, Derek was exceptionally productive and influential, impacting the field from concurrent prominent positions in academia and industry. At the time of his passing, Derek was a professor in the Chandra Family Department of Electrical and Computer Engineering at the University of Texas at Austin (UT Austin) and held the Temple Foundation Endowed Teaching Fellowship in Engineering. Derek was also a partner hardware architect of Microsoft Azure. Despite these demanding roles, Derek was always generous with his time to those around him. More than his technical accomplishments, all those who have crossed paths with Derek will remember him first as an unreservedly caring and giving colleague, collaborator, mentor, teacher, and friend.
- Open Access
- 1 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 178571-178586
Amorphica: A Fully Connected Annealer Supporting Metamorphic Annealing and Scalable Multi Chip Integration
- 1 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 178571-178586
Open AccessAmorphica: A Fully Connected Annealer Supporting Metamorphic Annealing and Scalable Multi Chip Integration
- Open Access
- 23 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 171458-171471
Efficient Stereo Visual Odometry on FPGA Featuring On-Chip Map Management and Pipelined Descriptor-Based Block Matching
Due to its ability to capture natural light without emitting additional signals, a stereo camera is a promising option for low-power visual odometry (VO). However, achieving real-time processing of VO using a stereo camera on embedded CPUs and GPUs is challenging due to the high computational demands and irregular computational patterns. These irregular patterns arise from the dynamic and non-uniform nature of feature detection, stereo matching, and map management, which vary significantly with scene complexity and camera movement. Various hardware accelerators have been proposed to address this challenge, but they typically tackle individual aspects of stereo VO, leading to difficulties in combining them efficiently. This paper introduces a comprehensive stereo VO accelerator implemented on an AMD Kria KV260 FPGA board, aimed at improving processing efficiency and reducing energy consumption. Our accelerator features innovative techniques including on-FPGA map management and pipelined stereo matching with the use of descriptors instead of image patches. These advancements result in significant reduction in off-chip data transfer, memory savings, and faster processing times. Additionally, an adaptive bucketing-based feature selection method is employed to enhance system accuracy without significant increases in hardware resource usage. Evaluation using the KITTI dataset shows that our accelerator achieves speedups of up to 3.08× and 2.75× over CPU-only and CPU+GPU implementations of LVT, an optimized algorithm, on an NVIDIA Jetson Nano developer kit B01 4GB, with corresponding energy efficiency improvements of 3.55× and 1.95×. The results demonstrate the accelerator’s effectiveness for real-time applications, highlighting the benefits of FPGA-based solutions in complex visual processing tasks.
- 23 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 171458-171471
Open AccessEfficient Stereo Visual Odometry on FPGA Featuring On-Chip Map Management and Pipelined Descriptor-Based Block Matching
Due to its ability to capture natural light without emitting additional signals, a stereo camera is a promising option for low-power visual odometry (VO). However, achieving real-time processing of VO using a stereo camera on embedded CPUs and GPUs is challenging due to the high computational demands and irregular computational patterns. These irregular patterns arise from the dynamic and non-uniform nature of feature detection, stereo matching, and map management, which vary significantly with scene complexity and camera movement. Various hardware accelerators have been proposed to address this challenge, but they typically tackle individual aspects of stereo VO, leading to difficulties in combining them efficiently. This paper introduces a comprehensive stereo VO accelerator implemented on an AMD Kria KV260 FPGA board, aimed at improving processing efficiency and reducing energy consumption. Our accelerator features innovative techniques including on-FPGA map management and pipelined stereo matching with the use of descriptors instead of image patches. These advancements result in significant reduction in off-chip data transfer, memory savings, and faster processing times. Additionally, an adaptive bucketing-based feature selection method is employed to enhance system accuracy without significant increases in hardware resource usage. Evaluation using the KITTI dataset shows that our accelerator achieves speedups of up to 3.08× and 2.75× over CPU-only and CPU+GPU implementations of LVT, an optimized algorithm, on an NVIDIA Jetson Nano developer kit B01 4GB, with corresponding energy efficiency improvements of 3.55× and 1.95×. The results demonstrate the accelerator’s effectiveness for real-time applications, highlighting the benefits of FPGA-based solutions in complex visual processing tasks.