Language：Others   Publishing type：Research paper (scientific journal)   Publisher：American Association for the Advancement of Science ({AAAS})  

Designing fluorescent molecules requires considering multiple interrelated molecular properties, as opposed to properties that straightforwardly correlated with molecular structure, such as light absorption of molecules. In this study, we have used a de novo molecule generator (DNMG) coupled with quantum chemical computation (QC) to develop fluorescent molecules, which are garnering significant attention in various disciplines. Using massive parallel computation (1024 cores, 5 days), the DNMG has produced 3643 candidate molecules. We have selected an unreported molecule and seven reported molecules and synthesized them. Photoluminescence spectrum measurements demonstrated that the DNMG can successfully design fluorescent molecules with 75% accuracy (
<italic>n</italic>
= 6/8) and create an unreported molecule that emits fluorescence detectable by the naked eye.

DOI： 10.1126/sciadv.abj3906

Web of Science

Scopus

PubMed

researchmap

Language：Others   Publishing type：Research paper (other academic)   Publisher：IEEE  

Data augmentation policies drastically improve the performance of image recognition tasks, especially when the policies are optimized for the target data and tasks. In this paper, we propose to optimize image recognition models and data augmentation policies simultaneously to improve the performance using gradient descent. Unlike prior methods, our approach avoids using proxy tasks or reducing search space, and can directly improve the validation performance. Our method achieves efficient and scalable training by approximating the gradient of policies by implicit gradient with Neumann series approximation. We demonstrate that our approach can improve the performance of various image classification tasks, including fine-grained image recognition, without using dataset-specific hyperparameter tuning.

DOI： 10.1109/WACV51458.2022.00359

Scopus

researchmap

Other Link： https://dblp.uni-trier.de/db/conf/wacv/wacv2022.html#HatayaZYN22

CiNii Research

Language：Others   Publishing type：Research paper (scientific journal)  

Language：English   Publishing type：Research paper (scientific journal)  

<title>Abstract</title>The nature of the recovery process of posttraumatic stress disorder (PTSD) symptoms is multifactorial. The Massive Parallel Limitless-Arity Multiple-testing Procedure (MP-LAMP), which was developed to detect significant combinational risk factors comprehensively, was utilized to reveal hidden combinational risk factors to explain the long-term trajectory of the PTSD symptoms. In 624 population-based subjects severely affected by the Great East Japan Earthquake, 61 potential risk factors encompassing sociodemographics, lifestyle, and traumatic experiences were analyzed by MP-LAMP regarding combinational associations with the trajectory of PTSD symptoms, as evaluated by the Impact of Event Scale-Revised score after eight years adjusted by the baseline score. The comprehensive combinational analysis detected 56 significant combinational risk factors, including 15 independent variables, although the conventional bivariate analysis between single risk factors and the trajectory detected no significant risk factors. The strongest association was observed with the combination of short resting time, short walking time, unemployment, and evacuation without preparation (adjusted <italic>P</italic> value = 2.2 × 10⁻⁴, and raw <italic>P</italic> value = 3.1 × 10⁻⁹). Although short resting time had no association with the poor trajectory, it had a significant interaction with short walking time (<italic>P</italic> value = 1.2 × 10⁻³), which was further strengthened by the other two components (<italic>P</italic> value = 9.7 × 10⁻⁵). Likewise, components that were not associated with a poor trajectory in bivariate analysis were included in every observed significant risk combination due to their interactions with other components. Comprehensive combination detection by MP-LAMP is essential for explaining multifactorial psychiatric symptoms by revealing the hidden combinations of risk factors.

DOI： 10.1038/s41598-020-78966-z

Language：English   Publishing type：Research paper (other academic)  

Modern shared-memory platforms embrace the Non-uniform Memory Access (NUMA) architecture - they have physically distributed, yet cache-coherent shared-memory. This paper explores the feasibility of a shared-memory graph processing engine for NUMA platforms inspired by designs that target zero-sharing platforms. This work exploits the characteristics of two processing modes, synchronous and asynchronous, in the context of the shared-memory NUMA platform. Depending on the algorithm, phase of execution, and graph topology, synchronous and asynchronous modes hold unique advantages over one another. We then explore a hybrid solution that combines synchronous and asynchronous processing within the same graph computation task and harness optimizations therein. An extensive evaluation using graphs with billions of edges and empirical comparisons with several state-of-the-art solutions demonstrate the performance advantages of our design.

DOI： 10.1109/bigdata50022.2020.9378430

Language：English   Publishing type：Research paper (scientific journal)  

Since the cost and time required to finetune parameters in traditional short-term traffic prediction models such as traffic simulators are very high, the prediction models have been developed mainly for managing recurrent congestion, rather than non-recurrent congestion caused, for example, by disaster. Machine learning models are promising candidates for traffic prediction during non-recurrent congestion due to their ability to tune parameters without a-priori knowledge, while their applicability to non-recurrent conditions has rarely been explored. To fill in this gap, this study conducts an exploratory analysis on the applicability of various machine learning models during a transportation network disruption with particular focuses on their ability to predict traffic states and the interpretability of the results. The analysis is conducted by using data obtained during the massive transport network disruption which occurred in Hiroshima in July 2018 due to heavy rain and subsequent landslides. The models tested include random forest, support vector machine, XGBoost, shallow feed-forward neural network, and deep feed-forward neural network. The results indicate that random forest and XGBoost methods produced the best results in terms of prediction accuracy. On the other hand, deep neural network models produce better results in terms of the interpretability of the results, i.e., the results can be logically explained from the perspective of existing traffic flow theory. These findings indicate that the model which produces the best prediction accuracy is not always the best for practical use since it does not mimic the mechanisms of congestion occurrence.

DOI： 10.1016/j.tranpol.2020.05.023

Language：English   Publishing type：Research paper (other academic)  

Data augmentation methods are indispensable heuristics to boost the performance of deep neural networks, especially in image recognition tasks. Recently, several studies have shown that augmentation strategies found by search algorithms outperform hand-made strategies. Such methods employ black-box search algorithms over image transformations with continuous or discrete parameters and require a long time to obtain better strategies. In this paper, we propose a differentiable policy search pipeline for data augmentation, which is much faster than previous methods. We introduce approximate gradients for several transformation operations with discrete parameters as well as a differentiable mechanism for selecting operations. As the objective of training, we minimize the distance between the distributions of augmented and original data, which can be differentiated. We show that our method, Faster AutoAugment, achieves significantly faster searching than prior methods without a performance drop.

DOI： 10.1007/978-3-030-58595-2_1

Language：English   Publishing type：Research paper (scientific journal)  

<title>Abstract</title>In computer-assisted synthesis planning (CASP) programs, providing as many chemical synthetic routes as possible is essential for considering optimal and alternative routes in a chemical reaction network. As the majority of CASP programs have been designed to provide one or a few optimal routes, it is likely that the desired one will not be included. To avoid this, an exact algorithm that lists possible synthetic routes within the chemical reaction network is required, alongside a recommendation of synthetic routes that meet specified criteria based on the chemist’s objectives. Herein, we propose a chemical-reaction-network-based synthetic route recommendation framework called “CompRet” with a mathematically guaranteed enumeration algorithm. In a preliminary experiment, CompRet was shown to successfully provide alternative routes for a known antihistaminic drug, cetirizine. CompRet is expected to promote desirable enumeration-based chemical synthesis searches and aid the development of an interactive CASP framework for chemists.

DOI： 10.1186/s13321-020-00452-5

Language：English   Publishing type：Research paper (scientific journal)  

DOI： 10.1080/14686996.2020.1793382

Language：Others   Publishing type：Research paper (other academic)  

Faster AutoAugment: Learning Augmentation Strategies Using Backpropagation.

DOI： 10.1007/978-3-030-58595-2_1

Language：Others   Publishing type：Research paper (scientific journal)  

Language：English   Publishing type：Research paper (scientific journal)  

DOI： 10.1093/bioinformatics/bty219

Language：Japanese  

Using local minima to accelerate Krawczyk-Hansen global optimization

Language：English   Publishing type：Research paper (other academic)  

An Extended GLB Library for Optimization Problems

Language：English   Publishing type：Research paper (scientific journal)  

Automatic design of organic materials requires black-box optimization in a vast chemical space. In conventional molecular design algorithms, a molecule is built as a combination of predetermined fragments. Recently, deep neural network models such as variational autoencoders and recurrent neural networks (RNNs) are shown to be effective in de novo design of molecules without any predetermined fragments. This paper presents a novel Python library ChemTS that explores the chemical space by combining Monte Carlo tree search and an RNN. In a benchmarking problem of optimizing the octanol-water partition coefficient and synthesizability, our algorithm showed superior efficiency in finding high-scoring molecules. ChemTS is available at https://github.com/tsudalab/ChemTS.[GRAPHICS].

DOI： 10.1080/14686996.2017.1401424

Language：English   Publishing type：Research paper (scientific journal)  

Background: Artificially synthesized RNA molecules provide important ways for creating a variety of novel functional molecules. State-of-the-art RNA inverse folding algorithms can design simple and short RNA sequences of specific GC content, that fold into the target RNA structure. However, their performance is not satisfactory in complicated cases.Result: We present a new inverse folding algorithm called MCTS-RNA, which uses Monte Carlo tree search (MCTS), a technique that has shown exceptional performance in Computer Go recently, to represent and discover the essential part of the sequence space. To obtain high accuracy, initial sequences generated by MCTS are further improved by a series of local updates. Our algorithm has an ability to control the GC content precisely and can deal with pseudoknot structures. Using common benchmark datasets for evaluation, MCTS-RNA showed a lot of promise as a standard method of RNA inverse folding.Conclusion: MCTS-RNA is available at https://github.com/tsudalab/MCTS-RNA.

DOI： 10.1186/s12859-017-1882-7

Language：English   Publishing type：Research paper (scientific journal)  

DOI： 10.1080/14686996.2017.1344083

Language：Others   Publishing type：Research paper (scientific journal)  

Language：Japanese  

Optimization of Playout Policy Integrated with Monte-Carlo Tree Search

Language：English   Publishing type：Research paper (other academic)  

We present a scalable parallel solver for numerical constraint satisfaction problems (NCSPs). Our parallelization scheme consists of homogeneous worker solvers, each of which runs on an available core and communicates with others via the global load balancing (GLB) method. The search tree of the branch and prune algorithm is split and distributed through the two phases of GLB: a random workload stealing phase and a workload distribution and termination phase based on a hyper-cube-shaped graph called lifeline. The parallel solver is simply implemented with X10 that provides an implementation of GLB as a library. In experiments, NCSPs from the literature were solved and attained up to 516-fold speedup using 600 cores of the TSUBAME2.5 supercomputer. Optimal GLB configurations are analyzed.

DOI： 10.1145/2771774.2771776

Language：English  

DOI： 10.1007/978-3-319-10428-7_30

Language：Japanese  

Humanlike AI for Mario Bros. Based on Evolutionary Computation and UCT

Language：English  

DOI： 10.1007/978-3-642-31866-5_1

Language：English   Publishing type：Research paper (other academic)  

After a computer named "Deep Blue" defeated the world chess champion Garry Kasparov in 1997, researchers studying computer board games focused their attention on the game "Go." Go is known to be more difficult for computers to play than chess or shogi because (1) the search space for Go is much larger, (2) it is difficult to define an appropriate evaluation function of position, and (3) a position sometimes changes globally in just one move. Recently, a new meth ad called Monte Carlo Go has been developed, which involves performing Monte Carlo simulations to evaluate a position. Monte Carlo Go increases the strength of the Computer-Go program. For Monte Carlo Go, the strength fully depends on the number of simulations. Several attempts were made to accelerate simulations, e.g., by the use of cluster systems and FPGAs. The cluster system yields good results, but it is a very expensive system. On the other hand, acceleration using an FPGA was not so easy because the usage of FPGA resources tends to be high. Previously, FPGA acceleration was feasible for smaller board such as a board with a 9 x 9 grid, while it was not feasible for the standard board with a 19 x 19 grid. In this paper, we propose triple line-based playout for Go (TLPG), a hardware algorithm for generating simulations using an FPGA. By reproducing global information redundantly, TLPG enables the generation of simulations only using local operations; this helps realize compact implementations of hardware logic, and thus, TLPG can handle both 9 x 9 and 19 x 19 grid Go boards. We implement TLPG on Xilinx Virtex-5 (XC5VFX70T-IFF1136) and evaluate it. TLPG can perform 40,649 playouts per second for a 9 x 9 grid Go board and 4,668 playouts per second for a 19 x 19 grid Go board.

DOI： 10.1109/reconfig.2009.75

Language：English   Publishing type：Research paper (other academic)  

We propose here a security evaluation methodology of image-based biometrics authentication systems against wolf attacks. A wolf attack is an attempt to impersonate a victim by feeding wolves into the system to be attacked. The wolf is input data that can be falsely accepted as a match with multiple templates. To create a secure system, we must evaluate the possibility of wolf attacks. Existing studies have relied on theoretical analysis of algorithms carried out by human beings, which is only effective if theoretical analysis is possible. Therefore, we propose a more generic approach based on a search to assist the developers of matching algorithms. We searched for wolves by using a recently developed algorithm called Monte-Carlo Tree Search (MCTS). We succeeded in detecting wolves in a matching algorithm, which appears promising considering that this is the first trial for this kind of approach. ©2009 IEEE.

DOI： 10.1109/btas.2009.5339016

Language：English  

AND-OR Tree Search Algorithms for Domains with Uniform Branching Factors

Language：English  

DOI： 10.1007/978-3-540-87608-3_13

Language：English  

DOI： 10.1007/978-3-540-88353-1_1

Language：Japanese  

Language：Japanese  

λ Search Based on Proof and Disproof Numbers
We present the df-pn λ search algorithm that combines threats with proof and disproof numbers. λ search is a promising method based on threats. Df-pn is an efficient algorithm that employs the notion of proof and disproof numbers. However, λ search uses neither proof nor disproof numbers, whereas df-pn incorporates no information on threat levels. Integrating threats with proof and disproof numbers is a natural extension to further enhance the search performance. We introduce pseudo-nodes for various threat levels at each node, to represent a node searched with a specific threat level. Then the proof and disproof numbers of the original node are defined using pseudo-nodes, which provides a model that can be searched with df-pn. We compared df-pn λ with df-pn on games with different properties. The results showed that df-pn λ is better than df-pn in Shogi and Go.

Language：Others  

Efficient Implementation of the Lambda Search based on Proof Numbers

Language：English   Publishing type：Research paper (other academic)  

Language：English  

Language：Japanese  

Language：English   Publishing type：Research paper (other academic)  

Speculative Parallel Execution on JVM

Event date： 2017.9

researchmap

Event date： 2015.9

Language：Japanese  

Country：Other  

Scalable Parallelization of a Numerical Constraint Solver

Event date： 2015.9

Language：Japanese  

researchmap

Event date： 2010.7

Language：Japanese  

Country：Other  

A Proposal for Distributed Parallel Monte Carlo Tree Search Framework
We propose to implement a framework for Monte Carlo Tree Search (MCTS) on distributed parallel systems. The objective is to facilitate the parallelization of existing sequential Monte Carlo Tree Search programs. It is expected to be effective for highly parallel distributed systems by using Transposition table Driven Scheduling as the parallelization technique for the framework.

Event date： 2010.7

Language：Japanese  

We propose to implement a framework for Monte Carlo Tree Search (MCTS) on distributed parallel systems. The objective is to facilitate the parallelization of existing sequential Monte Carlo Tree Search programs. It is expected to be effective for highly parallel distributed systems by using Transposition table Driven Scheduling as the parallelization technique for the framework.

researchmap

Event date： 2009.9

Language：Japanese  

Country：Other  

Event date： 2009.7

Language：Japanese  

Country：Other  

Acceleration of Monte-Carlo Go by FPGA-based Hardware
In the monte-carlo simulation of Go, it takes time to run playouts. There were attempts of accelerating by implementing circuits for playout on FPGA, but it is difficult to realize high-speed playouts because of high utilization of resources in a FPGA. In this paper, we propose an algorithm, Triple Line-based Playout for Go (TLPG) to accelerate playouts for the monte-carlo tree search for computer-go game. We implemented the playout logics on FPGA for 9x9 and 19x19 boards. With the optimizations, We achieved 13104playouts/sec in 9x9 and 2055playouts/sec in 19x19 board in simulation. By making games with GNU Go on a host Computer, We evaluation the playouts of TLPG.

Event date： 2009.7

Language：Japanese  

In the monte-carlo simulation of Go, it takes time to run playouts. There were attempts of accelerating by implementing circuits for playout on FPGA, but it is difficult to realize high-speed playouts because of high utilization of resources in a FPGA. In this paper, we propose an algorithm, Triple Line-based Playout for Go (TLPG) to accelerate playouts for the monte-carlo tree search for computer-go game. We implemented the playout logics on FPGA for 9x9 and 19x19 boards. With the optimizations, We achieved 13104playouts/sec in 9x9 and 2055playouts/sec in 19x19 board in simulation. By making games with GNU Go on a host Computer, We evaluation the playouts of TLPG.

researchmap

Event date： 2007.11

Language：Japanese  

Country：Other  

Df-pn Algorithm with Dynamic Widening

Event date： 2007.11

Language：Japanese  

researchmap

Event date： 2007.11

Language：Others  

Country：Other  

Depth-First UCT and Its Application to Go

Event date： 2007.11

Language：Others  

Country：Other  

Event date： 2007.11

researchmap

Event date： 2007.11

researchmap

Event date： 2005.9

Language：English  

Country：Other  

Searching for Double Threats in Subproblems of the Game of Go
It is difficult to make a fast and accurate evaluation function for the whole board in the Game of Go. Therefore sub-goal directed search is used widely among Go playing programs. One problem of sub-goal directed search is dependencies between sub-goals. There are several researches which aim to resolve the dependencies by obtaining the area which involves with the result of subgoals. An idea called relevancy zone is being used in some researches. In this paper, we introduce an algorithm which search for an area which would improve relevancy zone. The intersection of two such areas will be the candidate for double threat.

Event date： 1998.8

Language：Japanese  

Country：Other  

Implementing Parallel Speculative Execution of Loops on JVM
There have been several proposals about hardware speculative executions, in a larger gran-ularity than instruction level parallelism, by partitioning the target program into blocks.We have applied speculative execution onto Java Virtual Machine. We implemented it on a shared memory machine. The target for speculative execution is limited to loops. We measured speedups for simple loops and found that it is possible to gain speedups for loops which contains more than 10000 instructions by an interpreter Java Virtual Machine.

Event date： 1998.8

Language：Japanese  

There have been several proposals about hardware speculative executions, in a larger gran-ularity than instruction level parallelism, by partitioning the target program into blocks.We have applied speculative execution onto Java Virtual Machine. We implemented it on a shared memory machine. The target for speculative execution is limited to loops. We measured speedups for simple loops and found that it is possible to gain speedups for loops which contains more than 10000 instructions by an interpreter Java Virtual Machine.

researchmap

Language：Japanese  

Country：Japan  

Implementing Parallel Speculative Execution of Loops on JVM
There have been several proposals about hardware speculative executions, in a larger gran-ularity than instruction level parallelism, by partitioning the target program into blocks.We have applied speculative execution onto Java Virtual Machine. We implemented it on a shared memory machine. The target for speculative execution is limited to loops. We measured speedups for simple loops and found that it is possible to gain speedups for loops which contains more than 10000 instructions by an interpreter Java Virtual Machine.

Language：English  

Country：Japan  

Searching for Double Threats in Subproblems of the Game of Go(Session 3)
It is difficult to make a fast and accurate evaluation function for the whole board in the Game of Go. Therefore sub-goal directed search is used widely among Go playing programs. One problem of sub-goal directed search is dependencies between sub-goals. There are several researches which aim to resolve the dependencies by obtaining the area which involves with the result of subgoals. An idea called relevancy zone is being used in some researches. In this paper, we introduce an algorithm which search for an area which would improve relevancy zone. The intersection of two such areas will be the candidate for double threat.

Language：Japanese  

Country：Japan  

Df-pn Algorithm with Dynamic Widening

Language：Others  

Country：Japan  

Language：Others  

Country：Japan  

Depth-First UCT and Its Application to Go

Language：Japanese  

Country：Other  

Language：Japanese  

Country：Other  

Language：Japanese  

Country：Japan  

Acceleration of Monte-Carlo Go by FPGA-based Hardware
In the monte-carlo simulation of Go, it takes time to run playouts. There were attempts of accelerating by implementing circuits for playout on FPGA, but it is difficult to realize high-speed playouts because of high utilization of resources in a FPGA. In this paper, we propose an algorithm, Triple Line-based Playout for Go (TLPG) to accelerate playouts for the monte-carlo tree search for computer-go game. We implemented the playout logics on FPGA for 9x9 and 19x19 boards. With the optimizations, We achieved 13104playouts/sec in 9x9 and 2055playouts/sec in 19x19 board in simulation. By making games with GNU Go on a host Computer, We evaluation the playouts of TLPG.

Language：Japanese  

Country：Japan  

A Proposal for Distributed Parallel Monte Carlo Tree Search Framework
We propose to implement a framework for Monte Carlo Tree Search (MCTS) on distributed parallel systems. The objective is to facilitate the parallelization of existing sequential Monte Carlo Tree Search programs. It is expected to be effective for highly parallel distributed systems by using Transposition table Driven Scheduling as the parallelization technique for the framework.

Language：English  

Country：Other  

Language：Japanese  

Country：Japan  

Scalable Parallelization of a Numerical Constraint Solver

Language：Japanese  

Country：Japan  

Optimization of Playout Policy Integrated with Monte-Carlo Tree Search

Language：Others  

Country：Japan  

An Extended GLB Library for Optimization Problems

Language：Japanese  

Country：Japan  

Using local minima to accelerate Krawczyk-Hansen global optimization

Language：Others  

Country：Japan  

Language：Others  

Country：Japan  

Language：Japanese  

Efficient Processing Techniques for Very Large-scale Graph Structure

Language：Japanese  

「ゲーム木検索」／「スピントロニクス」／「金星のスーパーローテーション」／「グラフェン」／「獲得免疫と自然免疫」／（あとがき）

Language：Japanese  

Language：Japanese  

AND-OR Tree Search Algorithms for Domains with Uniform Branching Factors(Foundations of AI,<Special Issue>Doctorial Theses on Aritifical Intelligence)

DOI： 10.11517/jjsai.25.1_148

Language：Japanese  

Monte-Carlo Tree Search : A Revolutionary Algorithm Developed for Computer Go