Satoru Hiwa

Hybrid optimization using DIRECT, GA, and SQP for global exploration

This paper presents a new hybrid optimization approach, which combines multiple optimization algorithms. To develop an efficient hybrid optimization algorithm, it is necessary to determine how the optimization process is performed. This paper focuses on the balance between local and broad searches, and multiple optimization methods are controlled to derive both the optimum point and the information of the landscape. By this approach, we can describe the global landscape after derivation of optimization. To achieve the proposed optimization strategy, three distinguished optimization algorithms are introduced: DIRECT (Dividing RECTangles), GAs (Genetic Algorithms), and SQP (Sequential Quadratic Programming). To integrate these three algorithms, each algorithm, especially DIRECT, was modified and developed. The performance of the proposed hybrid algorithm was examined through numerical experiments. From these experiments, not only the optimum point but also the information of the landscape was determined. The information of the landscape verified the reliability of optimization results.

Novel search scheme for multi-objective evolutionary algorithms to obtain well-approximated and widely spread Pareto solutions

Abstract In multi-objective optimization, the quality of Pareto-optimal solutions is evaluated by the efficiency of the optimal front (proximity), uniformity, and spread. This paper introduces a novel search scheme for multi-objective evolutionary algorithms (MOEAs), whose solutions demonstrate improved proximity and spread metrics. Our proposed scheme comprises two search phases with different search objectives. The first phase uses a reference-point-based approach to improve proximity; the second phase adopts a distributed-cooperation scheme (DC-scheme) to broaden the range of solutions. We experimentally investigate the effectiveness of our proposed scheme on the walking fish group (WFG) test suite of scalable multi-objective problems. Finally, we show the applicability of the proposed scheme to various types of MOEAs.

Simulated Annealing using an Adaptive Search Vector

It is reported that simulated annealing (SA), which changes one design variable at a time, is effective when applied to high-dimensional continuous optimization problems. However, if a correlation exists among the design variables, it is not efficient to search each dimension. In this paper, we propose SA with a mechanism to determine an appropriate search direction according to the landscape of the given problems. Its effectiveness is verified for high-dimensional problems with correlation among design variables

Design Mode Analysis of Pareto Solution Set for Decision-Making Support

Decision making in engineering design problems is challenging because they have multiple and conflicting criteria and complex correlation between design parameters. This study proposes a decision-making support methodology named design mode analysis, which consists of data clustering and principal component analysis (PCA). A design mode is indicated by the eigenvector obtained by PCA and reveals the dominant design parameters in a given dataset. The proposed method is a general framework to obtain the design modes from high-dimensional and large datasets. The effectiveness of the proposed method is verified on the conceptual design problem of the hybrid rocket engine.

Validity of decision mode analysis on an ROI determination problem in multichannel fNIRS data

Region of interest (ROI) determination is necessary when using functional near-infrared spectroscopy (fNIRS) data to detect brain activity. To extract ROIs from multiple fNIRS channels, we investigated the validity of applying decision mode analysis to the fNIRS dataset. This classifies a dataset into clusters with similar features. For each cluster, the dataset is decomposed into a mean vector and a linear combination of eigenvectors. Applying this to fNIRS signals, the mean vector can be used to represent change in hemoglobin (Hb), and the eigenvectors interpreted as a signal component constructing the arbitrary signal. Characterizing these vectors by correlating them with a theoretical model of brain function aids our understanding of where Hb changes occur and what type of Hb changes reflect brain activity in fNIRS data. Decision mode analysis of fNIRS data measured during viewing stereoscopic images identified ROIs around the right inferior frontal gyrus associated with attentional control, and frontal association area associated with decision on action and prediction. Our experimental results showed that information obtained from decision mode analysis can aid quantitative and qualitative ROI determination.

Region-of-interest extraction of fMRI data using genetic algorithms

Functional connectivity, which is indicated by time-course correlations of brain activities among different brain regions, is one of the most useful metrics to represent human brain states. In functional connectivity analysis (FCA), the whole brain is parcellated into a certain number of regions based on anatomical atlases, and the mean time series of brain activities are calculated. Then, the correlation between mean signals of two regions is repeatedly calculated for all combinations of regions, and finally, we obtain the correlation matrix of the whole brain. FCA allows us to understand which regions activate cooperatively during specific stimulus or tasks. In this study, we attempt to represent human brain states using functional connectivity as feature vectors. As there are a number of brain regions, it is difficult to determine which regions are prominent to represent the brain state. Therefore, we proposed an automatic region-of-interest (ROI) extraction method to classify human brain states. Time-series brain activities were measured by functional magnetic resonance imaging (fMRI), and FCA was performed. Each element of the correlation matrix was used as a feature vector for brain state classification, and element characteristics were learned using supervised learning methods. The elements used as feature vectors, i.e., ROIs, were determined automatically using a genetic algorithm to maximize the classification accuracy of brain states. fMRI data measured during two emotional conditions, i.e., pleasant and unpleasant emotions, were used to show the effectiveness of the proposed method. Numerical experiments revealed that the proposed method could extract the superior frontal gyrus, orbitofrontal cortex, cuneus, cerebellum, and cerebellar vermis as ROIs associated with pleasant and unpleasant emotions.

Reference point-based search scheme for multiobjective evolutionary algorithm

In multiobjective optimization problems, it is important to derive solutions with high accuracy, uniform distribution, and broadness. Here, we propose a two-phase search process to improve the accuracy and broadness of solutions. In the first phase, the accuracy of the solutions is improved, and a reference point specified by a decision maker is set and utilized for the search. In the second phase, the solutions are broadened using the Distributed Cooperation Scheme. The numerical results indicated that the proposed search scheme was capable of deriving broader solutions than the conventional multiobjective evolutionary algorithm without deterioration of accuracy.

Solution exploration using multi-objective genetic algorithm for determining experiment candidate

Solving problems involves the following two phases. In the first phase, detail of the problem are determined and the solution is found according to these condition. In the second phase, the results are used to narrow down any remaining ambiguities in the problem. In terms of the flow of a river, the former lies upstream of solving the problem while the latter lies downstream. Multiobjective genetic algorithms (GAs) is are able to find possible solution sets involving trade-offs among several different objective functions. In this study, we use a multiobjective GA to grasp variety types of solutions, not solve the problem directly. In other words, we use it for the upstream problem-solving step. As a case study of using multiobjective GAs to explore solutions, we identify cancer cells where the Nrf3 transcription factor is active and consider the problem of determining which genes to focus on in experiments based on that information. In this case, we selected gene candidates that are likely to be associated with Nrf3 activity and experiments (which previously had to be carried out exhaustively) are currently being carried out to confirm these results.

Automated Extraction of Human Functional Brain Network Properties Associated with Working Memory Load through a Machine Learning-Based Feature Selection Algorithm

Working memory (WM) load-dependent changes of functional connectivity networks have previously been investigated by graph theoretical analysis. However, the extraordinary number of nodes represented within the complex network of the human brain has hindered the identification of functional regions and their network properties. In this paper, we propose a novel method for automatically extracting characteristic brain regions and their graph theoretical properties that reflect load-dependent changes in functional connectivity using a support vector machine classification and genetic algorithm optimization. The proposed method classified brain states during 2- and 3-back test conditions based upon each of the three regional graph theoretical metrics (degree, clustering coefficient, and betweenness centrality) and automatically identified those brain regions that were used for classification. The experimental results demonstrated that our method achieved a >90% of classification accuracy using each of the three graph metrics, whereas the accuracy of the conventional manual approach of assigning brain regions was only 80.4%. It has been revealed that the proposed framework can extract meaningful features of a functional brain network that is associated with WM load from a large number of nodal graph theoretical metrics without prior knowledge of the neural basis of WM.

Functional near-infrared spectroscopy study of the neural correlates between auditory environments and intellectual work performance

Many people spend a considerable amount of time performing intellectual activities within auditory environments that affect work efficiency. To investigate auditory environments that improve working efficiency, we investigated the relationship between brain activity and performance of the number memory task in environments with and without white noise using functional near‐infrared spectroscopy (fNIRS).