Data augmentation is one of the most effective ways to stabilize learning by improving the generalization of machine-learning models. In recent years, automatic data augmentation methods, such as AutoAugment or Fast AutoAugment have been attracting attention; and these methods improved the results of image classification and object detection tasks. However, several problems remain. Most notably, a larger training dataset requires higher computational costs. When searching with a small dataset in an attempt to determine the data augmentation approach, the true data space and sampling data space do not fully correspond with each other, thereby causing the generalization performance to deteriorate. Moreover, in the existing automatic augmentation methods, the search phase is often dominated by an exceptional sub-policy, which results in a loss of diversity of operations. In this study, we solved these problems by introducing evolutionary computation to previous methods. As mentioned earlier, maintaining diversity is essential. Therefore, we adopted the thermodynamical genetic algorithm (TDGA), which can control the population diversity with a specific genetic operator, known as the thermodynamical selection rule. To confirm the effectiveness of the proposed method, computational experiments were conducted using two benchmark datasets, CIFAR-10 and SVHN, as examples. The experimental results show that the proposed method can obtain various useful augmentation sub-policies for the problems while reducing the computational cost.

Although areas devastated by natural disasters must be promptly reconstructed, reconstruction works involving construction machinery are dangerous and bring a significant risk of a second disaster. A remote construction system is thus often used, which enables an operator to conduct reconstruction work from a remote and safe place while viewing on-site videos. However, the remote construction has a major problem: it has much lower work efficiency than manned construction because videos are delayed and impaired in a wireless network environment. In this paper, we present a remote construction system using our proposed adaptive video streaming method that dynamically adjusts the video bitrate on the basis of the predicted communication throughput variation. The experiments show that the proposed remote construction system using adaptive video streaming reduces operation time by 25.5%, which corresponds to a 34.3% improvement in work efficiency.

In the construction industry, remote construction has been introduced in specified cases such as restoration work after natural disasters in order to improve safety. Remote construction can also contribute to allocating limited labor resources efficiently and solving the labor scarcity problem. For this, long-distance remote construction though networks such as a wireless local area network (WLAN) and the Internet must be made possible. However, long-distance networks can have large transmission delays and packet loss, which cause bottlenecks that result in low operability and increased workload. We propose a remote construction system based on adaptive remote control techniques for improving video streaming and operation with considerations made for delays and packet loss. We also conduct an experiment to evaluate how the proposed remote construction system reduces workloads. Our experiment shows that the proposed system reduces the number of manipulations for a certain task by 25%. It also reduces the duration of operations by 16%, which corresponds to a 20% improvement in work efficiency.

Recently, deep learning has been studied as one of the most effective methods in the machine learning field, and lots of results have been reported. However, the most effective way to construct neural networks has not yet been determined. Besides, interpretation of an obtained network is difficult. In a previous study, we proposed a novel method to construct a neural network using a support vector machine called SVM-NN. However, there is also a problem in that, for hard to apply to a nonlinear problem and model size problem. In this study, we first propose a new network structure called AND/OR layers to solve the nonlinear problem of SVM-NN. AND/OR layers improve the effectiveness of identification by grouping support vectors based on training data results. This type of novel SVM-NN is called SVM-NN(AND/OR). We also utilize the genetic algorithm to pair and reduce support vectors to compress the model size of SVM-NN and SVM-NN(AND/OR). To confirm the effectiveness of the proposed methods, the computational experiments were carried out taking typical benchmark problems as examples. The effectiveness of the proposed method is confirmed by computer simulations.

Genetic Programmings (GPs) is one of the most powerful evolutionary computation (EC) for software evolution. In ECs, it is difficult to maintain efficient building blocks. In particular, the control of building blocks in the population of genetic programming (GP) is relatively difficult because of tree-shaped individuals and also because of bloat, which is the uncontrolled growth of ineffective code segments in GP. For a variety of reasons, reliable techniques to remove bloat are highly desirable. This paper introduces a novel approach of removing bloat, by proposing a novel GP called “Genetic Programming with Multi-Layered Population Structure (MLPS-GP)” that employs multi-layered population and searches solutions using local search and crossover. The MLPS-GP has no mutation-like operator because such kinds of operators are the source of bloats. We showed that diversity can be maintained well only controlling the tree structures by a well-structured multi-layered population. To confirm the effectiveness of the proposed method, the computational experiments were carried out taking several classical Boolean problems as examples.

In this paper we try to realize a system that enables users to interact with surrounding surfaces using touch interactions. For this purpose, we propose new touch detection technique which utilizes the shadows of a finger, and developed a prototype system with an infrared (IR) camera and two IR lights. Since the shapes of a finger's shadows vary drastically depending on the distance between the surface and finger, our prototype system can detect touch. To improve the accuracy of the estimated touch position, we also introduce multiple regression analysis into the estimation algorithm of the touch position. We conducted two experiments on the accuracy of estimated touch position, and the results shows that the target accuracy was within an error of less than 5 mm.

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.

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.

Many evolutionary techniques such as genetic algorithms (GAs) employ parameters that facilitate user control of search dynamics. However, these parameters require time-consuming tuning processes to avoid problems such as premature convergence. 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. P3 approaches to search would be useful for constructing a novel theory for optimal search using GAs. In this study, we propose a novel technique "analysis based on the Distribution of Inferior Individuals in the local neighborhood" (DII analysis). DII analysis can be applied to local search techniques, including P3. The computational complexity of applied problems can be estimated based on a number of local optima according to the results obtained using DII. We also propose combining P3 with DII analysis (P3-DII), which controls the maximum number of fitness evaluations performed by genetic operators. The computational experiments were carried out taking several combinational problems as examples. According to our experimental results, we demonstrated that P3-DII found several optimal solutions that P3 failed to find.

Shin et al [19] and McKay et al [15] previously applied tree compression and semantics-based simplification to study the distribution of building blocks in evolving Genetic Programming populations. However their method could only give static estimates of the degree of repetition of building blocks in one generation at a time, supplying no information about the flow of building blocks between generations. Here, we use a state-of-the-art tree compression algorithm, xmlppm, to estimate the extent to which frequent building blocks from one generation are still in use in a later generation. While they compared the behaviour of different GP algorithms on one specific problem -- a simple symbolic regression problem -- we extend the analysis to a more complex problem, a symbolic regression problem to find a Fourier approximation to a sawtooth wave, and to a Boolean domain, odd parity.

Compression algorithms generate a predictive model of data, using the model to reduce the number of bits required to transmit the data (in effect, transmitting only the differences from the model). As a consequence, the degree of compression achieved provides an estimate of the level of regularity in the data. Previous work has investigated the use of these estimates to understand the replication of building blocks within genetic programming (GP) individuals, and hence to understand how different GP algorithms promote the evolution of repeated common structure within individuals. Here, we extend this work to the population level, and use it to understand the extent of similarity between sub-structures within individuals in GP populations.

Genetic Algorithms (GAs) are a search and optimization technique based on the mechanism of evolution. Recently, another sort of population-based optimization method called Estimation of Distribution Algorithms (EDAs) have been proposed to solve the GA's defects. Although several comparison studies between GAs and EDAs have been made, little is known about differences of statistical features between them. In this paper, we propose new statistical indices which are based on the concepts of crossover and mutation, used in GAs, to analyze the behavior of the population based optimization techniques. We also show simple results of comparison studies between GAs and the Bayesian Optimization Algorithm (BOA), a well-known Estimation of Distribution Algorithms (EDAs).

Adaptation to dynamic environments is an important application of genetic algorithms (GAs). However, there are many difficulties to apply the GA to dynamic environments. Especially, in online environments, the GA's defects become remarkable because individuals should be evaluated in the real world. We proposes a novel approach to such an online adaptation called the environment identifying genetic algorithm (EIGA). Computer simulation is carried out by taking an Nk-landscape problem as an example.

In this paper, a finishing system with a mounted abrasive tool is proposed for polishing of metallic molds. The shape of the mounted abrasive tool is a ball-end type. When a metallic mold with curved surface is polished, not only the orientation of the mounted abrasive tool is fixed but also its revolution is locked. The motion of the mounted abrasive tool is feedforward controlled based on an initial trajectory calculated in advance. The surface is polished by the polishing force which consists of a contact force, motion and viscous friction forces acting between the mold and tool. In this case, the velocities in the normal and tangent directions are delicately regulated so that the polishing force can track the desired value. The effectiveness of the proposed system is proved by experiments using an industrial robot with a PC based controller.

An implementation of a biped robot which is capable of dynamic walking by a simple nonlinear control algorithm is presented. Four DC servo motors actuate the knee and ankle joints of the legs of the robot. The biped is constrained to the sagital plane, and the motion generation is reduced to a problem of controlling the position and velocity of the robot's center of gravity. They are controlled by a nonlinear feedback controller, based on a sample feedback linearization method. Several design issues including mechanical structure, leg actuation, and control system of the robot are discussed. Experimental results demonstrate the effectiveness of the algorithm.