Taku Hasegawa
Osaka Prefecture University
Network
Latest external collaboration on country level. Dive into details by clicking on the dots.
Publication
Featured researches published by Taku Hasegawa.
genetic and evolutionary computation conference | 2014
Taku Hasegawa; Kaname Matsumura; Kaiki Tsuchie; Naoki Mori; Keinosuke Matsumoto
Introducing the machine learning technique into evolutionary computation (EC) is one of the most important issues to expand EC design. In this paper, we proposed a novel method that combines the genetic algorithm and support vector machine to achieve the imaginary evolution without real fitness evaluations.
Archive | 2017
Kento Tsukada; Taku Hasegawa; Naoki Mori; Keinosuke Matsumoto
One of the most important issues for evolutionary computation (EC) is to consider the number of fitness evaluations. In order to reduce the number of fitness evaluations, we have proposed the novel surrogate model called Rank Space Estimation (RSE) model and the surrogate-assisted EC with RSE model called the Fitness Landscape Learning Evolutionary Computation (FLLEC). This paper presents a novel CMA-ES with RSE model for continuous optimization problems and a scaling method for input data to surrogate model.
Archive | 2017
Saya Fujino; Taku Hasegawa; Miki Ueno; Naoki Mori; Keinosuke Matsumoto
Creating interactive picture books based on human “Kansei” is one of the most interesting and difficult issues in the artificial intelligence field. We have proposed a novel interactive picture book based on Pictgent (Picture Information Shared Conversation Agent) and CASOOK (Creative Animating Sketchbook). Since our system accepts human sketches instead of natural languages, a high degree of sketch recognition accuracy is required. Recently, convolutional neural networks (CNNs) have been applied to various image- recognition tasks successfully. We have also adopted a CNN model for the sketch recognition of the proposed interactive picture book. However, it takes a considerable effort to tune the hyperparameters of a CNN. In this paper, we propose a novel parameter tuning method for CNNs using an evolutionary approach. The effectiveness of the proposed method is confirmed by a computer simulation that uses, as an example, a scribble-object recognition problem for the interactive picture book.
international symposium on distributed computing | 2018
Ryo Iwasaki; Taku Hasegawa; Naoki Mori; Keinosuke Matsumoto
Deep learning has developed into one of the most powerful methods in the machine learning field. In particular, convolutional neural networks (CNNs) have been applied not only to image recognition tasks but also to natural language processing (NLP). To reuse older deep learning models, transfer learning techniques have been widely used in the image recognition field. However, there has been little research on transfer learning in NLP. In this paper, we propose a novel transfer learning model based on a relaxation method of CNNs for NLP. The effectiveness of the proposed method is verified using computer simulations, taking a film review score recognition task as an example.
genetic and evolutionary computation conference | 2017
Naoki Mori; Taku Hasegawa; Kento Tsukada; Keinosuke Matsumoto
In order to reduce the number of fitness evaluations, the novel surrogate model called Rank Space Estimation (RSE) model and the surrogate-assisted EC with RSE model called the Fitness Landscape Learning Evolutionary Computation (FLLEC) have been proposed. In this paper, we analyze the scarling effect for CMA-ES with RSE model with support vector machine(SVM). The performance of CMA-ES with RSE model by using adequate scarling is shown by computer simulation taking k-table problem as an example.
Archive | 2017
Taku Hasegawa; Yuta Araki; Naoki Mori; Keinosuke Matsumoto
Unlike many GAs, the Parameter-less Population Pyramid (P3) is an optimization model that avoids premature convergence due to the pyramid-like structure of populations, and thus P3 can be applied to a wide range of problems without parameter tuning. However, in some problems, P3 cannot control the number of fitness evaluations in local search and in crossover, while adapting problem structures. Meanwhile, we have proposed a novel technique, called DII analysis. The computational complexity of applied problems can be estimated based on the number of local optima according to the results obtained using DII. In order to solve the problem of P3, we also have proposed combining P3 with DII analysis (P3-DII). In this study, we investigated the effect of DII analysis on balance between genetic operators. The performance of P3-DII was confirmed according to the computational experiments which were carried out taking several combinational problems as examples.
genetic and evolutionary computation conference | 2015
Taku Hasegawa; Kento Tsukada; Naoki Mori; Keinosuke Matsumoto
Surrogate-assisted Evolutionary Computation provides us good results in real-world optimization. In this paper, we propose a novel adaptive evolution control using P-I similarity index for surrogate-assisted EC. The computational experiments are carried out to show the effectiveness of the proposed adaptive evolution control.
congress on evolutionary computation | 2015
Kazuyuki Inoue; Taku Hasegawa; Naoki Mori; Keinosuke Matsumoto
The optimal Exploration Exploitation Trade-off (EE Trade-off) is a fundamental goal in the field of Evolutionary Computation. To achieve the goal, we have proposed P-I similarity index and Dictyostelium based Genetic Algorithm (DGA). P-I similarity index provides an exploitation degree to enable applications to explicitly control EE Trade-off. DGA has specific operators which adopt the life cycle of dictyostelium to trade off between exploration and exploitation. In this study we specify the feature of P-I similarity index and introduce DGA with P-I similarity index. The computational experiments were carried out taking several combinatorial optimization problems as examples to suggest that DGA with P-I similarity index has wide applicability to discrete problems.
congress on evolutionary computation | 2015
Taku Hasegawa; Tsukada Kento; Naoki Mori; Keinosuke Matsumoto
Fitness approximation methods in Evolutionary Computation (EC) provide us good results in real-world optimization. On the other hand, little is known about the advantages and disadvantages of each surrogate models. Moreover, the performance of models depends on a structure of original function. Therefore, various kinds of surrogate models can leads to better results. We also have proposed a novel surrogate model which can estimate the only rank of two individuals using Support Vector Machine. In addition, we have proposed EC framework with that model called Fitness Landscape Learning Evolutionary Computation (FLLEC) which has shown good performance. In this paper, we compared two type of evolution control in FLLEC with the computational experiments.
genetic and evolutionary computation conference | 2015
Kazuyuki Inoue; Taku Hasegawa; Yuta Araki; Naoki Mori; Keinosuke Matsumoto