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

Reservoir computing (RC), first applied to temporal signal processing, is a recurrent neural network in which neurons are randomly connected. Once initialized, the connection strengths remain unchanged. Such a simple structure turns RC into a non-linear dynamical system that maps low-dimensional inputs into a high-dimensional space. The model's rich dynamics, linear separability, and memory capacity then enable a simple linear readout to generate adequate responses for various applications. RC spans areas far beyond machine learning, since it has been shown that the complex dynamics can be realized in various physical hardware implementations and biological devices. This yields greater flexibility and shorter computation time. Moreover, the neuronal responses triggered by the model's dynamics shed light on understanding brain mechanisms that also exploit similar dynamical processes. While the literature on RC is vast and fragmented, here we conduct a unified review of RC's recent developments from machine learning to physics, biology, and neuroscience. We first review the early RC models, and then survey the state-of-the-art models and their applications. We further introduce studies on modeling the brain's mechanisms by RC. Finally, we offer new perspectives on RC development, including reservoir design, coding frameworks unification, physical RC implementations, and interaction between RC, cognitive neuroscience and evolution.

Language：English   Publishing type：Research paper (international conference proceedings)  

Language：English   Publishing type：Research paper (international conference proceedings)  

Language：English   Publishing type：Research paper (international conference proceedings)  

Language：English   Publishing type：Research paper (international conference proceedings)  

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

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

Language：English   Publishing type：Research paper (international conference proceedings)  

Language：English   Publishing type：Research paper (international conference proceedings)  

Language：English   Publishing type：Research paper (international conference proceedings)  

Language：English   Publishing type：Research paper (international conference proceedings)  

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

Decision-making models in the behavioral, cognitive, and neural sciences typically consist of forced-choice paradigms with two alternatives. While theoretically it is feasible to translate any decision situation to a sequence of binary choices, real-life decision-making is typically more complex and nonlinear, involving choices among multiple items, graded judgments, and deferments of decision-making. Here, we discuss how the complexity of real-life decision-making can be addressed using conventional decision-making models by focusing on the interactive dynamics between criteria settings and the collection of evidence. Decision-makers can engage in multi-stage, parallel decision-making by exploiting the space for deliberation, with non-binary readings of evidence available at any point in time. The interactive dynamics principally adhere to the speed-accuracy tradeoff, such that increasing the space for deliberation enables extended data collection. The setting of space for deliberation reflects a form of meta-decision-making that can, and should be, studied empirically as a value-based exercise that weighs the prior propensities, the economics of information seeking, and the potential outcomes. Importantly, the control of the space for deliberation raises a question of agency. Decision-makers may actively and explicitly set their own decision parameters, but these parameters may also be set by environmental pressures. Thus, decision-makers may be influenced—or nudged in a particular direction—by how decision problems are framed, with a sense of urgency or a binary definition of choice options. We argue that a proper understanding of these mechanisms has important practical implications toward the optimal usage of space for deliberation.

Language：English   Publishing type：Research paper (international conference proceedings)  

Language：English   Publishing type：Research paper (international conference proceedings)  

Language：English   Publishing type：Research paper (international conference proceedings)  

Language：English  

Language：English  

Language：English  

Language：English  

Language：English  

Language：English  

Language：English  

Adversarial machine learning has indicated that perturbations to a picture may disable a deep neural network from correctly qualifying the content of a picture. The progressing research has even revealed that the perturbations do not necessarily have to be large in size. This research has been transplanted to traffic signs. The test results were disastrous. For example, a perturbated stop sign was recognized as a speeding sign. Because visualization technology is not able to overcome this problem yet, the question arises who should be liable for accidents caused by this technology. Manufacturers are being pointed at and for that reason it has been claimed that the commercialization of autonomous vehicles may stall. Without autonomous vehicles, the sharing economy may not fully develop either. This chapter shows that there are alternatives for the unpredictable financial burden on the car manufacturers for accidents with autonomous cars. This chapter refers to operator liability, but argues that for reasons of fairness, this is not a viable choice. A more viable choice is a no-fault liability on the manufacturer, as this kind of scheme forces the car manufacturer to be careful but keeps the financial risk predicable. Another option is to be found outside law. Engineers could build infrastructure enabling automation. Such infrastructure may overcome the problems of the visualization technology, but could potentially create a complex web of product and service providers. Legislators should prevent that the victims of an accident, if it were still to occur, would face years in court with the various actors of this complex web in order to receive compensation.

DOI： 10.1007/978-981-15-1350-3_11

Language：English  

Other Link： https://www.springer.com/gp/book/9789811536847

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

Recent research has revealed that the output of deep neural networks (DNNs) can be easily altered by adding relatively small perturbations to the input vector. In this paper, we analyze an attack in an extremely limited scenario where only one pixel can be modified. For that we propose a novel method for generating one-pixel adversarial perturbations based on differential evolution (DE). It requires less adversarial information (a black-box attack) and can fool more types of networks due to the inherent features of DE. The results show that 67.97% of the natural images in Kaggle CIFAR-10 test dataset and 16.04% of the ImageNet (ILSVRC 2012) test images can be perturbed to at least one target class by modifying just one pixel with 74.03% and 22.91% confidence on average. We also show the same vulnerability on the original CIFAR-10 dataset. Thus, the proposed attack explores a different take on adversarial machine learning in an extreme limited scenario, showing that current DNNs are also vulnerable to such low dimension attacks. Besides, we also illustrate an important application of DE (or broadly speaking, evolutionary computation) in the domain of adversarial machine learning: creating tools that can effectively generate low-cost adversarial attacks against neural networks for evaluating robustness.

DOI： 10.1109/TEVC.2019.2890858

Language：English   Publishing type：Research paper (international conference proceedings)  

Language：English   Publishing type：Research paper (international conference proceedings)  

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

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

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

Language：English   Publishing type：Research paper (international conference proceedings)  

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

The output of convolutional neural networks (CNNs) has been shown to be discontinuous which can make the CNN image classifier vulnerable to small well-tuned artificial perturbation. That is, images modified by conducting such alteration (i.e., adversarial perturbation) that make little difference to the human eyes can completely change the CNN classification results. In this paper, we propose a practical attack using differential evolution (DE) for generating effective adversarial perturbations. We comprehensively evaluate the effectiveness of different types of DEs for conducting the attack on different network structures. The proposed method only modifies five pixels (i.e., few-pixel attack), and it is a black-box attack which only requires the miracle feedback of the target CNN systems. The results show that under strict constraints which simultaneously control the number of pixels changed and overall perturbation strength, attacking can achieve 72.29%, 72.30%, and 61.28% non-targeted attack success rates, with 88.68%, 83.63%, and 73.07% confidence on average, on three common types of CNNs. The attack only requires modifying five pixels with 20.44, 14.28, and 22.98 pixel value distortion. Thus, we show that current deep neural networks are also vulnerable to such simpler black-box attacks even under very limited attack conditions.

DOI： 10.1186/s41074-019-0053-3

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

Moving Target Defense (MTD) concept has been a feasible idea for reducing the possibility of attack happening through alternation attack surfaces or diversification the attribute or parameters of a protected system. As a result of applying MTD techniques to the system, an attacker would have more difficulties in exploiting a vulnerabilities of the target system. This study proposes an evaluation method of MTD systems combined with several different MTD techniques. The proposed method is a primary step in designing an evaluation model for the effectiveness of MTD. The main goal is to estimate the attack success ratio on the MTD systems mitigating from threats of executable binary file or malware injection. With the proposed evaluation method, we expect to prove that the MTD technology can enhance the security of a web server, and can be applied in a real-world information system. As our preliminary work done, we set up a prototype framework to validate the proposed work in a pseudo-experimental environment.

DOI： 10.1109/CANDARW.2018.00111

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

In 2021, it is anticipated that there will be approximately 30 billion Internet of Things (IoT) devices. The tremendous aggregate value of the IoT makes it a tempting and lucrative target for cyber criminals. The breakout of Mirai malware, which compromises poorly secured IoT devices with factory-default username and passphrase to launch Distributed Denial of Service (DDoS) attacks, has raised broad awareness towards the need for increased IoT security. To better defend against Mirai infection and spread, it is critical to know how the malware operates as the first step. In this paper, we give a combined static and dynamic analysis of Mirai, basing on the results of which, we introduce the application of Threat Tracer. Threat tracer is an information system simulator initially developed to help design a system robust against Advanced Persistent Attacks(APT). It offers an intuitive track on how a cyber threat behaves in a complicated networked system. The feedback simultaneously contributes to revealing vulnerabilities of a system. Our work focuses on the replication of Mirai Malware's operating processes in Threat Tracer simulation. By achieving doing so, we believe it could offer a comprehensible description of how Mirai acts. Also, considering the continuous emergence of Mirai variants, the simulation serves as a predictor on upcoming threats' behavior patterns.

DOI： 10.1109/CANDARW.2018.00104

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

Modern cryptographic schemes is developed based on the mathematical theory. Recently works show a new direction about cryptography based on the neural networks. Instead of learning a specific algorithm, a cryptographic scheme is generated automatically. While one kind of neural network is used to achieve the scheme, the idea of the neural cryptography can be realized by other neural network architecture is unknown. In this paper, we make use of this property to create neural cryptography scheme on a new topology evolving neural network architecture called Spectrum-diverse unified neuroevolution architecture. First, experiments are conducted to verify that Spectrum-diverse unified neuroevolution architecture is able to achieve automatic encryption and decryption. Subsequently, we do experiments to achieve the neural symmetric cryptosystem by using adversarial training.

DOI： 10.1109/CANDARW.2018.00091

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

Recent research has shown that deep convolutional neural networks (DCNN) are vulnerable to several different types of attacks while the reasons of such vulnerability are still under investigation. For instance, the adversarial perturbations can conduct a slight change on a natural image to make the target DCNN make the wrong recognition, while the reasons that DCNN is sensitive to such small modification are divergent from one research to another. In this paper, we evaluate the robustness of two commonly used activation functions of DCNN, namely the sigmoid and ReLu, against the recently proposed low-dimensional one-pixel attack. We show that the choosing of activation functions can be an important factor that influences the robustness of DCNN. The results show that comparing with sigmoid, the ReLu non-linearity is more vulnerable which allows the low dimensional one-pixel attack exploit much higher success rate and confidence of launching the attack. The results give insights on designing new activation functions to enhance the security of DCNN.

DOI： 10.1109/CANDARW.2018.00049

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

DOI： 10.1145/3205651.3207869

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

The Internet of Things (IoT) is an extension of the traditional Internet, which allows a very large number of smart devices, such as home appliances, network cameras, sensors and controllers to connect to one another to share information and improve user experiences. IoT devices are micro-computers for domain-specific computations rather than traditional functionspecific embedded devices. This opens the possibility of seeing many kinds of existing attacks, traditionally targeted at the Internet, also directed at IoT devices. As shown by recent events, such as the Mirai and Brickerbot botnets, DDoS attacks have become very common in IoT environments as these lack basic security monitoring and protection mechanisms. In this paper, we propose a novel light-weight approach for detecting DDos malware in IoT environments. We extract the malware images (i.e., a one-channel gray-scale image converted from a malware binary) and utilize a light-weight convolutional neural network for classifying their families. The experimental results show that the proposed system can achieve 94:0% accuracy for the classification of goodware and DDoS malware, and 81:8% accuracy for the classification of goodware and two main malware families.

DOI： 10.1109/COMPSAC.2018.10315

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

Third-party tracking, which can collect the users' privacy when users are surfing the Internet, has garnered much attention. Nowadays tracker-blocking tools often use a ruleset based on the domains and elements that need to be blocked. This results in blocking all access tracking, even though the website shows no sign about tracking users' privacy. And what's more, although the tracker-blocking tools try their best to block all the third-party tracking, not all the users dislike the advertisement. Some of them think if their privacy is fine, it's all right to accept advertisements. In this paper, we present a novel framework by using Word2Vec to block third-party tracking. Our goal is to create more flexible and well-developed ruleset that can help users to protect their privacy according to their needs. Instead of blocking all access tracking, we decide to pay more attention to the websites that have a strong probability to collect the users' privacy. We use Word2Vec to classify the websites, and our results show that after using our framework, the error rate drops from 71% to 24%. We believe it brings the new blood into the field of web privacy by providing not only the new third-party tracking tool but also a novel way of thinking about how to block the third-party tracking.

DOI： 10.1109/CANDAR.2017.26

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

Learning algorithms are being increasingly adopted in various applications. However, further expansion will require methods that work more automatically. To enable this level of automation, a more powerful solution representation is needed. However, by increasing the representation complexity, a second problem arises. The search space becomes huge, and therefore, an associated scalable and efficient searching algorithm is also required. To solve both the problems, first a powerful representation is proposed that unifies most of the neural networks features from the literature into one representation. Second, a new diversity preserving method called spectrum diversity is created based on the new concept of chromosome spectrum that creates a spectrum out of the characteristics and frequency of alleles in a chromosome. The combination of spectrum diversity with a unified neuron representation enables the algorithm to either surpass or equal NeuroEvolution of Augmenting Topologies on all of the five classes of problems tested. Ablation tests justify the good results, showing the importance of added new features in the unified neuron representation. Part of the success is attributed to the novelty-focused evolution and good scalability with a chromosome size provided by spectrum diversity. Thus, this paper sheds light on a new representation and diversity preserving mechanism that should impact algorithms and applications to come.

DOI： 10.1109/TNNLS.2016.2551748

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

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

In the domain of information security, code obfuscation is a feature often employed for malicious purposes. For example there have been quite a few papers reporting that obfuscated JavaScript frequently comes with malicious functionality such as redirecting to external malicious websites. In order to capture such obfuscation, a class of detectors based on statistical features of code, mostly n-grams have been proposed and been claimed to achieve high detection accuracy. In this paper, we formalize a common scenario between defenders who maintain the statistical obfuscation detectors and adversaries who want to evade the detection. Accordingly, we create two kinds of evasion attack methods and evaluate the robustness of statistical detectors under such attacks. Experimental results show that statistical obfuscation detectors can be easily fooled by a sophisticated adversary even in worst case scenarios.

DOI： 10.1007/978-3-319-64200-0_8

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

Intrinsic motivation and novelty search are promising approaches to deal with plateaus, deceptive functions and other exploration problems where using only the main objective function is insufficient. However, it is not clear until now how and if intrinsic motivation (novelty search) can improve single objective algorithms in general. The hurdle is that using multi-objective algorithms to deal with single-objective problems adds an unnecessary overhead such as the search for non-dominated solutions. Here, we propose the Curious algorithm which is the first multi-objective algorithm focused on solving single-objective problems. Curious uses two subpopulations algorithms. One subpopulation is dedicated for improving objective function values and another one is added to search for unknown regions of space based on objective prediction errors. By using a differential evolution operator, genes from individuals in all subpopulations are mixed. In this way, the promising regions (solutions with high fitness) and unknown regions (solutions with high prediction error) are searched simultaneously. Because of thus realized strong yet well controlled novelty search, the algorithm possesses powerful exploration ability and outperforms usual single population based algorithms such as differential evolution. Thus, it demonstrates that the addition of intrinsic motivation is promising and should improve further single objective algorithms in general.

DOI： 10.1145/2908961.2908982

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

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

In the real world, the environment is constantly changing with the input variables under the effect of noise. However, few algorithms were shown to be able to work under those circumstances. Here, Novelty-Organizing Team of Classifiers (NOTC) is applied to the continuous action mountain car as well as two variations of it: A noisy mountain car and an unstable weather mountain car. These problems take respectively noise and change of problem dynamics into account. Moreover, NOTC is compared with NeuroEvolution of Augmenting Topologies (NEAT) in these problems, revealing a trade-off between the approaches. While NOTC achieves the best performance in all of the problems, NEAT needs less trials to converge. It is demonstrated that NOTC achieves better performance because of its division of the input space (creating easier problems). Unfortunately, this division of input space also requires a bit of time to bootstrap.

DOI： 10.1109/CEC.2015.7257254

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

Problems can be categorized as fractured or unfractured ones. A different set of characteristics are needed for learning algorithms to solve each of these two types of problems. However, the exact characteristics needed to solve each type are unclear. This article shows that the division of the input space is one of these characteristics. In other words, a study is presented showing that while fractured problems benefit from a finer division of the input space, unfractured problems benefit from a coarser division of input space. Many open questions still remains. And the article discusses two conjectures which can be used to solve fractured problems more easily.

DOI： 10.1145/2739482.2768447

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

Structured evolutionary algorithms have been investigated for some time. However, they have been under explored especially in the field of multi-objective optimization. Despite good results, the use of complex dynamics and structures keep the understanding and adoption rate of structured evolutionary algorithms low. Here, we propose a general subpopulation framework that has the capability of integrating optimization algorithms without restrictions as well as aiding the design of structured algorithms. The proposed framework is capable of generalizing most of the structured evolutionary algorithms, such as cellular algorithms, island models, spatial predator-prey, and restricted mating based algorithms. Moreover, we propose two algorithms based on the general subpopulation framework, demonstrating that with the simple addition of a number of single-objective differential evolution algorithms for each objective, the results improve greatly, even when the combined algorithms behave poorly when evaluated alone at the tests. Most importantly, the comparison between the subpopulation algorithms and their related panmictic algorithms suggests that the competition between different strategies inside one population can have deleterious consequences for an algorithm and reveals a strong benefit of using the subpopulation framework.

DOI： 10.1162/EVCO_a_00118

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

In reinforcement learning, there are basically two spaces to search: value-function space and policy space. Consequently, there are two fitness functions each with their associated trade-offs. However, the problem is still perceived as a single-objective one. Here a multi-objective reinforcement learning algorithm is proposed with a structured novelty map population evolving feedforward neural models. It outperforms a gradient based continuous input-output state-of-art algorithm in two problems. Contrary to the gradient based algorithm, the proposed one solves both problems with the same parameters and smaller variance of results. Moreover, the results are comparable even with other discrete action algorithms of the literature as well as neuroevolution methods such as NEAT. The proposed method brings also the novelty map population concept, i.e., a novelty map-based population which is less sensitive to the input distribution and therefore more suitable to create the state space. In fact, the novelty map framework is shown to be less dynamic and more resource efficient than variants of the self-organizing map.

DOI： 10.1109/SICE.2014.6935299

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

It is widely known that reinforcement learning is a more general problem than supervised learning. In fact, supervised learning can be seen as a class of reinforcement learning problems. However, only a couple of papers tested reinforcement learning algorithms in supervised learning problems. Here we propose a new and simpler way to abstract supervised learning for any reinforcement learning algorithm. Moreover, a new algorithm called Novelty-Organizing Classifiers is developed based on a Novelty Map population that focuses more on the novelty of the inputs than their frequency. A comparison of the proposed method with Self-Organizing Classifiers and BioHel on some datasets is presented. Even though BioHel is specialized in solving supervised learning problems, the results showed only a trade-off between the algorithms. Lastly, results on a maze problem validate the flexibility of the proposed algorithm beyond supervised learning problems. Thus, Novelty-Organizing Classifiers is capable of solving many supervised learning problems as well as a maze problem without changing any parameter at all. Considering the fact that no adaptation of parameters was executed, the proposed algorithm's basis seems interestingly flexible.

DOI： 10.1145/2598394.2598429

Language：Others  

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

Learning classifier systems (LCSs) are evolutionary machine learning algorithms, flexible enough to be applied to reinforcement, supervised and unsupervised learning problems with good performance. Recently, self organizing classifiers were proposed which are similar to LCSs but have the advantage that in its structured population no balance between niching and fitness pressure is necessary. However, more tests and analysis are required to verify its benefits. Here, a variation of the first algorithm is proposed which uses a parameterless self organizing map (SOM). This algorithm is applied in challenging problems such as big, noisy as well as dynamically changing continuous input-action mazes (growing and compressing mazes are included) with good performance. Moreover, a genetic operator is proposed which utilizes the topological information of the SOM's population structure, improving the results. Thus, the first steps in structured evolutionary machine learning are shown, nonetheless, the problems faced are more difficult than the state-of-art continuous input-action multi-step ones.

DOI： 10.1007/s12065-013-0095-x

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

Learning classifier systems are adaptive learning systems which have been widely applied in a multitude of application domains. However, there are still some generalization problems unsolved. The hurdle is that fitness and niching pressures are difficult to balance. Here, a new algorithm called Self Organizing Classifiers is proposed which faces this problem from a different perspective. Instead of balancing the pressures, both pressures are separated and no balance is necessary. In fact, the proposed algorithm possesses a dynamical population structure that self-organizes itself to better project the input space into a map. The niched fitness concept is defined along with its dynamical population structure, both are indispensable for the understanding of the proposed method. Promising results are shown on two continuous multi-step problems. One of which is yet more challenging than previous problems of this class in the literature.

DOI： 10.1145/2463372.2463501

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

Language：English  

This paper presents a new technique for optimizing binary problems with building blocks. The authors have developed a different approach to existing Estimation of Distribution Algorithms (EDAs). Our technique, called Phylogenetic Differential Evolution (PhyDE), combines the Phylogenetic Algorithm and the Differential Evolution Algorithm. The first one is employed to identify the building blocks and to generate metavariables. The second one is used to find the best instance of each metavariable. In contrast to existing EDAs that identify the related variables at each iteration, the presented technique finds the related variables only once at the beginning of the algorithm, and not through the generations. This paper shows that the proposed technique is more efficient than the well known EDA called Extended Compact Genetic Algorithm (ECGA), especially for large-scale systems which are commonly found in real world problems.

DOI： 10.4018/978-1-4666-4253-9.ch002

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

Ensembles of classifiers have been studied for some time. It is widely known that weak learners should be accurate and diverse. However, in the real world there are many constraints and few have been said about the robustness of ensembles and how to develop it. In the context of ran- dom subspace methods, this paper addresses the question of developing ensembles to face problems under time con- straints. Experiments show that selecting weak learners based on their accuracy can be used to create robust en- sembles. Thus, the selection pressure in ensembles is a key technique to create not just effective ensembles but also robust ones. Moreover, the experiments motivate further research on ensembles made of low dimensional classifiers which achieve general accurate results.

DOI： 10.1145/2464576.2480775

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

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

Learning classifier systems have been solving reinforcement learning problems for some time. However, they face difficulties under multi-step continuous problems. Adaptation may also become harder with time since the convergence of the population decreases its diversity. This article demonstrate that the novel Self Organizing Classifiers method can cope with dynamical multi-step continuous problems. Moreover, adaptation remains the same after convergence.

DOI： 10.1109/DevLrn.2013.6652558

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

When applied to high-dimensional datasets, feature selection algorithms might still leave dozens of irrelevant variables in the dataset. Therefore, even after feature selection has been applied, classifiers must be prepared to the presence of irrelevant variables. This paper investigates a new training method called Contingency Training which increases the accuracy as well as the robustness against irrelevant attributes. Contingency training is classifier independent. By subsampling and removing information from each sample, it creates a set of constraints. These constraints aid the method to automatically find proper importance weights of the dataset's features. Experiments are conducted with the contingency training applied to neural networks over traditional datasets as well as datasets with additional irrelevant variables. For all of the tests, contingency training surpassed the unmodified training on datasets with irrelevant variables and even outperformed slightly when only a few or no irrelevant variables were present.

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

In general, Multi-objective Evolutionary Algorithms do not guarantee find solutions in the Pareto-optimal set. We propose a new approach for solving decomposable deceptive multi-objective problems that can find all solutions of the Pareto-optimal set. Basically, the proposed approach starts by decomposing the problem into subproblems and, then, combining the found solutions. The resultant approach is a Multi-objective Estimation of Distribution Algorithm for solving relatively complex multi-objective decomposable problems, using a probabilistic model based on a phylogenetic tree. The results show that, for the tested problem, the algorithm can efficiently find all the solutions of the Pareto-optimal set, with better scaling than the hierarchical Bayesian Optimization Algorithm and other algorithms of the state of art.

DOI： 10.1007/978-3-642-19893-9_20

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

Event date： 2020.6

Language：English   Presentation type：Symposium, workshop panel (public)  

Country：Japan  

Event date： 2019.11

Language：English  

Venue：Nagasaki   Country：Japan  

Generating image and video is a hot topic in Deep Learning. Especially, generating video is a difficult but meaningful work. How to generate video which has diversity and plausibility is still a problem to be solved. In this paper, we propose a novel model of Generative Adversarial Network(GAN) which called FDGAN to generate clear contour lines. Unlike existing GAN that only use frames, our method extends to use inter-frame difference. First introduce two temporal difference methods to process the inter-frame. Then increase a frame difference discriminator to discriminate whether the inter-frame is true or not. Using the model and new structure proposed, we perform video generation experiments on several widely used benchmark datasets such as MOVING MNIST, UCF-101. Consequently, the results achieve state-of-the-art performance for clarifying contour lines. Both quantitative and qualitative evaluations were made to show the effectiveness of our methods.

Event date： 2019.9

Language：English  

Venue：Hiroshima   Country：Japan  

Intrinsically, driving is a Markov Decision Process which suits well the reinforcement learning paradigm. In this paper, we propose a novel agent which learns to drive a vehicle without any human assistance. We use the concept of reinforcement learning and evolutionary strategies to train our agent in a 2D simulation environment. Our model's architecture goes beyond the World Model's by introducing difference images in the autoencoder. This novel involvement of difference images in the auto-encoder gives a better representation of the latent space concerning the motion of the vehicle and helps an autonomous agent to learn more efficiently how to drive a vehicle. Results show that our method requires fewer (96% less) total agents, (87.5% less) agents per generations, (70% less) generations and (90% less) rollouts than the original architecture while achieving the same accuracy of the original.

Event date： 2019.7

Language：English  

Venue：Prague   Country：Czech Republic  

Lexicase selection achieves very good solution quality by introducing ordered test cases. However, the computational complexity of lexicase selection can prohibit its use in many applications. In this paper, we introduce Batch Tournament Selection (BTS), a hybrid of tournament and lexicase selection which is approximately one order of magnitude faster than lexicase selection while achieving a competitive quality of solutions. Tests on a number of regression datasets show that BTS compares well with lexicase selection in terms of mean absolute error while having a speed-up of up to 25 times. Surprisingly, BTS and lexicase selection have almost no difference in both diversity and performance. This reveals that batches and ordered test cases are completely different mechanisms which share the same general principle fostering the specialization of individuals. This work introduces an efficient algorithm that sheds light onto the main principles behind the success of lexicase, potentially opening up a new range of possibilities for algorithms to come.

Event date： 2019.7

Language：English  

Venue：Prague   Country：Czech Republic  

Modern gradient based optimization methods for deep neural networks demonstrate outstanding results on image classification tasks. However, methods that do not rely on gradient feedback fail to tackle deep network optimization. In the field of evolutionary computation, applying evolutionary algorithms directly to network weights remains to be an unresolved challenge. In this paper we examine a new framework for the evolution of deep nets. Based on the empirical analysis, we propose the use of linear sub-spaces of problems to search for promising optimization trajectories in parameter space, opposed to weight evolution. We show that linear sub-spaces of loss functions are sufficiently well-behaved to allow trajectory evaluation. Furthermore, we introduce fitness measure to show that it is possible to correctly categorize trajectories according to their distance from the optimal path. As such, this work introduces an alternative approach to evolutionary optimization of deep networks.

Event date： 2019.6

Language：English  

Venue：Wellington   Country：New Zealand  

Recently, deep learning based natural language processing techniques are being extensively used to deal with spam mail, censorship evaluation in social networks, among others. However, there is only a couple of works evaluating the vulnerabilities of such deep neural networks. Here, we go beyond attacks to investigate, for the first time, universal rules, i.e., rules that are sample agnostic and therefore could turn any text sample in an adversarial one. In fact, the universal rules do not use any information from the method itself (no information from the method, gradient information or training dataset information is used), making them black-box universal attacks. In other words, the universal rules are sample and method agnostic. By proposing a coevolutionary optimization algorithm we show that it is possible to create universal rules that can automatically craft imperceptible adversarial samples (only less than five perturbations which are close to misspelling are inserted in the text sample). A comparison with a random search algorithm further justifies the strength of the method. Thus, universal rules for fooling networks are here shown to exist. Hopefully, the results from this work will impact the development of yet more sample and model agnostic attacks as well as their defenses.

Event date： 2018.11

Language：English  

Venue：Takayama   Country：Japan  

Moving Target Defense (MTD) concept has been a feasible idea for reducing the possibility of attack happening through alternation attack surfaces or diversification the attribute or parameters of a protected system. As a result of applying MTD techniques to the system, an attacker would have more difficulties in exploiting a vulnerabilities of the target system. This study proposes an evaluation method of MTD systems combined with several different MTD techniques. The proposed method is a primary step in designing an evaluation model for the effectiveness of MTD. The main goal is to estimate the attack success ratio on the MTD systems mitigating from threats of executable binary file or malware injection. With the proposed evaluation method, we expect to prove that the MTD technology can enhance the security of a web server, and can be applied in a real-world information system. As our preliminary work done, we set up a prototype framework to validate the proposed work in a pseudo-experimental environment.

Event date： 2018.11

Language：English  

Venue：Takayama   Country：Japan  

In 2021, it is anticipated that there will be approximately 30 billion Internet of Things (IoT) devices. The tremendous aggregate value of the IoT makes it a tempting and lucrative target for cyber criminals. The breakout of Mirai malware, which compromises poorly secured IoT devices with factory-default username and passphrase to launch Distributed Denial of Service (DDoS) attacks, has raised broad awareness towards the need for increased IoT security. To better defend against Mirai infection and spread, it is critical to know how the malware operates as the first step. In this paper, we give a combined static and dynamic analysis of Mirai, basing on the results of which, we introduce the application of Threat Tracer. Threat tracer is an information system simulator initially developed to help design a system robust against Advanced Persistent Attacks(APT). It offers an intuitive track on how a cyber threat behaves in a complicated networked system. The feedback simultaneously contributes to revealing vulnerabilities of a system. Our work focuses on the replication of Mirai Malware's operating processes in Threat Tracer simulation. By achieving doing so, we believe it could offer a comprehensible description of how Mirai acts. Also, considering the continuous emergence of Mirai variants, the simulation serves as a predictor on upcoming threats' behavior patterns.

Event date： 2018.11

Language：English  

Venue：Takayama   Country：Japan  

Modern cryptographic schemes is developed based on the mathematical theory. Recently works show a new direction about cryptography based on the neural networks. Instead of learning a specific algorithm, a cryptographic scheme is generated automatically. While one kind of neural network is used to achieve the scheme, the idea of the neural cryptography can be realized by other neural network architecture is unknown. In this paper, we make use of this property to create neural cryptography scheme on a new topology evolving neural network architecture called Spectrum-diverse unified neuroevolution architecture. First, experiments are conducted to verify that Spectrum-diverse unified neuroevolution architecture is able to achieve automatic encryption and decryption. Subsequently, we do experiments to achieve the neural symmetric cryptosystem by using adversarial training.

Event date： 2018.11

Language：English  

Venue：Takayama   Country：Japan  

Recent research has shown that deep convolutional neural networks (DCNN) are vulnerable to several different types of attacks while the reasons of such vulnerability are still under investigation. For instance, the adversarial perturbations can conduct a slight change on a natural image to make the target DCNN make the wrong recognition, while the reasons that DCNN is sensitive to such small modification are divergent from one research to another. In this paper, we evaluate the robustness of two commonly used activation functions of DCNN, namely the sigmoid and ReLu, against the recently proposed low-dimensional one-pixel attack. We show that the choosing of activation functions can be an important factor that influences the robustness of DCNN. The results show that comparing with sigmoid, the ReLu non-linearity is more vulnerable which allows the low dimensional one-pixel attack exploit much higher success rate and confidence of launching the attack. The results give insights on designing new activation functions to enhance the security of DCNN.

Event date： 2018.7

Language：English  

Venue：Tokyo   Country：Japan  

The Internet of Things (IoT) is an extension of the traditional Internet, which allows a very large number of smart devices, such as home appliances, network cameras, sensors and controllers to connect to one another to share information and improve user experiences. IoT devices are micro-computers for domain-specific computations rather than traditional functionspecific embedded devices. This opens the possibility of seeing many kinds of existing attacks, traditionally targeted at the Internet, also directed at IoT devices. As shown by recent events, such as the Mirai and Brickerbot botnets, DDoS attacks have become very common in IoT environments as these lack basic security monitoring and protection mechanisms. In this paper, we propose a novel light-weight approach for detecting DDos malware in IoT environments. We extract the malware images (i.e., a one-channel gray-scale image converted from a malware binary) and utilize a light-weight convolutional neural network for classifying their families. The experimental results show that the proposed system can achieve 94:0% accuracy for the classification of goodware and DDoS malware, and 81:8% accuracy for the classification of goodware and two main malware families.

Event date： 2018.7

Language：English  

Venue：Kyoto   Country：Japan  

Event date： 2017.11

Language：English  

Venue：Aomori   Country：Japan  

Third-party tracking, which can collect the users' privacy when users are surfing the Internet, has garnered much attention. Nowadays tracker-blocking tools often use a ruleset based on the domains and elements that need to be blocked. This results in blocking all access tracking, even though the website shows no sign about tracking users' privacy. And what's more, although the tracker-blocking tools try their best to block all the third-party tracking, not all the users dislike the advertisement. Some of them think if their privacy is fine, it's all right to accept advertisements. In this paper, we present a novel framework by using Word2Vec to block third-party tracking. Our goal is to create more flexible and well-developed ruleset that can help users to protect their privacy according to their needs. Instead of blocking all access tracking, we decide to pay more attention to the websites that have a strong probability to collect the users' privacy. We use Word2Vec to classify the websites, and our results show that after using our framework, the error rate drops from 71% to 24%. We believe it brings the new blood into the field of web privacy by providing not only the new third-party tracking tool but also a novel way of thinking about how to block the third-party tracking.

Event date： 2017.8 - 2017.9

Language：English  

Venue：Hiroshima   Country：Japan  

In the domain of information security, code obfuscation is a feature often employed for malicious purposes. For example there have been quite a few papers reporting that obfuscated JavaScript frequently comes with malicious functionality such as redirecting to external malicious websites. In order to capture such obfuscation, a class of detectors based on statistical features of code, mostly n-grams have been proposed and been claimed to achieve high detection accuracy. In this paper, we formalize a common scenario between defenders who maintain the statistical obfuscation detectors and adversaries who want to evade the detection. Accordingly, we create two kinds of evasion attack methods and evaluate the robustness of statistical detectors under such attacks. Experimental results show that statistical obfuscation detectors can be easily fooled by a sophisticated adversary even in worst case scenarios.

Event date： 2016.7

Language：English  

Venue：Denver   Country：United States  

Intrinsic motivation and novelty search are promising approaches to deal with plateaus, deceptive functions and other exploration problems where using only the main objective function is insufficient. However, it is not clear until now how and if intrinsic motivation (novelty search) can improve single objective algorithms in general. The hurdle is that using multi-objective algorithms to deal with single-objective problems adds an unnecessary overhead such as the search for non-dominated solutions. Here, we propose the Curious algorithm which is the first multi-objective algorithm focused on solving single-objective problems. Curious uses two subpopulations algorithms. One subpopulation is dedicated for improving objective function values and another one is added to search for unknown regions of space based on objective prediction errors. By using a differential evolution operator, genes from individuals in all subpopulations are mixed. In this way, the promising regions (solutions with high fitness) and unknown regions (solutions with high prediction error) are searched simultaneously. Because of thus realized strong yet well controlled novelty search, the algorithm possesses powerful exploration ability and outperforms usual single population based algorithms such as differential evolution. Thus, it demonstrates that the addition of intrinsic motivation is promising and should improve further single objective algorithms in general.

Event date： 2015.7

Language：English  

Venue：Madrid   Country：Spain  

Problems can be categorized as fractured or unfractured ones. A different set of characteristics are needed for learning algorithms to solve each of these two types of problems. However, the exact characteristics needed to solve each type are unclear. This article shows that the division of the input space is one of these characteristics. In other words, a study is presented showing that while fractured problems benefit from a finer division of the input space, unfractured problems benefit from a coarser division of input space. Many open questions still remains. And the article discusses two conjectures which can be used to solve fractured problems more easily.

Event date： 2015.5

Language：English  

Venue：Sendai   Country：Japan  

In the real world, the environment is constantly changing with the input variables under the effect of noise. However, few algorithms were shown to be able to work under those circumstances. Here, Novelty-Organizing Team of Classifiers (NOTC) is applied to the continuous action mountain car as well as two variations of it: A noisy mountain car and an unstable weather mountain car. These problems take respectively noise and change of problem dynamics into account. Moreover, NOTC is compared with NeuroEvolution of Augmenting Topologies (NEAT) in these problems, revealing a trade-off between the approaches. While NOTC achieves the best performance in all of the problems, NEAT needs less trials to converge. It is demonstrated that NOTC achieves better performance because of its division of the input space (creating easier problems). Unfortunately, this division of input space also requires a bit of time to bootstrap.

Event date： 2014.9

Language：English  

Venue：Sapporo   Country：Japan  

In reinforcement learning, there are basically two spaces to search: value-function space and policy space. Consequently, there are two fitness functions each with their associated trade-offs. However, the problem is still perceived as a single-objective one. Here a multi-objective reinforcement learning algorithm is proposed with a structured novelty map population evolving feedforward neural models. It outperforms a gradient based continuous input-output state-of-art algorithm in two problems. Contrary to the gradient based algorithm, the proposed one solves both problems with the same parameters and smaller variance of results. Moreover, the results are comparable even with other discrete action algorithms of the literature as well as neuroevolution methods such as NEAT. The proposed method brings also the novelty map population concept, i.e., a novelty map-based population which is less sensitive to the input distribution and therefore more suitable to create the state space. In fact, the novelty map framework is shown to be less dynamic and more resource efficient than variants of the self-organizing map.

Event date： 2014.7

Language：English  

Venue：Vancouver, BC   Country：Canada  

It is widely known that reinforcement learning is a more general problem than supervised learning. In fact, supervised learning can be seen as a class of reinforcement learning problems. However, only a couple of papers tested reinforcement learning algorithms in supervised learning problems. Here we propose a new and simpler way to abstract supervised learning for any reinforcement learning algorithm. Moreover, a new algorithm called Novelty-Organizing Classifiers is developed based on a Novelty Map population that focuses more on the novelty of the inputs than their frequency. A comparison of the proposed method with Self-Organizing Classifiers and BioHel on some datasets is presented. Even though BioHel is specialized in solving supervised learning problems, the results showed only a trade-off between the algorithms. Lastly, results on a maze problem validate the flexibility of the proposed algorithm beyond supervised learning problems. Thus, Novelty-Organizing Classifiers is capable of solving many supervised learning problems as well as a maze problem without changing any parameter at all. Considering the fact that no adaptation of parameters was executed, the proposed algorithm's basis seems interestingly flexible.

Event date： 2013.9

Language：English  

Venue：Nagoya   Country：Japan  

When applied to high-dimensional datasets, feature selection algorithms might still leave dozens of irrelevant variables in the dataset. Therefore, even after feature selection has been applied, classifiers must be prepared to the presence of irrelevant variables. This paper investigates a new training method called Contingency Training which increases the accuracy as well as the robustness against irrelevant attributes. Contingency training is classifier independent. By subsampling and removing information from each sample, it creates a set of constraints. These constraints aid the method to automatically find proper importance weights of the dataset's features. Experiments are conducted with the contingency training applied to neural networks over traditional datasets as well as datasets with additional irrelevant variables. For all of the tests, contingency training surpassed the unmodified training on datasets with irrelevant variables and even outperformed slightly when only a few or no irrelevant variables were present.

Event date： 2013.8

Language：English  

Venue：Osaka   Country：Japan  

Learning classifier systems have been solving reinforcement learning problems for some time. However, they face difficulties under multi-step continuous problems. Adaptation may also become harder with time since the convergence of the population decreases its diversity. This article demonstrate that the novel Self Organizing Classifiers method can cope with dynamical multi-step continuous problems. Moreover, adaptation remains the same after convergence.

Event date： 2013.7

Language：English  

Venue：Amsterdam   Country：Netherlands  

Learning classifier systems are adaptive learning systems which have been widely applied in a multitude of application domains. However, there are still some generalization problems unsolved. The hurdle is that fitness and niching pressures are difficult to balance. Here, a new algorithm called Self Organizing Classifiers is proposed which faces this problem from a different perspective. Instead of balancing the pressures, both pressures are separated and no balance is necessary. In fact, the proposed algorithm possesses a dynamical population structure that self-organizes itself to better project the input space into a map. The niched fitness concept is defined along with its dynamical population structure, both are indispensable for the understanding of the proposed method. Promising results are shown on two continuous multi-step problems. One of which is yet more challenging than previous problems of this class in the literature.

Event date： 2013.7

Language：English  

Venue：Amsterdam   Country：Netherlands  

Ensembles of classifiers have been studied for some time. It is widely known that weak learners should be accurate and diverse. However, in the real world there are many constraints and few have been said about the robustness of ensembles and how to develop it. In the context of ran- dom subspace methods, this paper addresses the question of developing ensembles to face problems under time con- straints. Experiments show that selecting weak learners based on their accuracy can be used to create robust en- sembles. Thus, the selection pressure in ensembles is a key technique to create not just effective ensembles but also robust ones. Moreover, the experiments motivate further research on ensembles made of low dimensional classifiers which achieve general accurate results.

Event date： 2011.4

Language：English  

Venue：Ouro Preto   Country：Brazil  

In general, Multi-objective Evolutionary Algorithms do not guarantee find solutions in the Pareto-optimal set. We propose a new approach for solving decomposable deceptive multi-objective problems that can find all solutions of the Pareto-optimal set. Basically, the proposed approach starts by decomposing the problem into subproblems and, then, combining the found solutions. The resultant approach is a Multi-objective Estimation of Distribution Algorithm for solving relatively complex multi-objective decomposable problems, using a probabilistic model based on a phylogenetic tree. The results show that, for the tested problem, the algorithm can efficiently find all the solutions of the Pareto-optimal set, with better scaling than the hierarchical Bayesian Optimization Algorithm and other algorithms of the state of art.

Language：Others  

Country：Other  

Language：Others  

Country：Other  

Language：Others  

Country：Other  

Language：Others  

Country：Other  

Language：Others  

Country：Other  

Language：Others  

Country：Other  

Language：Others  

Country：Other  

Language：Others  

Country：Other  

Language：Others  

Country：Other  

Language：Others  

Country：Other  

Language：Others  

Country：Other  

Language：Others  

Country：Other  

Language：Others  

Country：Other  

Language：Others  

Country：Other  

Language：Others  

Country：Other  

Type：Competition, symposium, etc. 

Type：Peer review 

Number of peer-reviewed articles in foreign language journals：10

Number of peer-reviewed articles in Japanese journals：0

Proceedings of International Conference Number of peer-reviewed papers：10

Proceedings of domestic conference Number of peer-reviewed papers：0

Type：Competition, symposium, etc. 

Type：Competition, symposium, etc. 

Type：Peer review 

Number of peer-reviewed articles in foreign language journals：20

Number of peer-reviewed articles in Japanese journals：0

Proceedings of International Conference Number of peer-reviewed papers：20

Proceedings of domestic conference Number of peer-reviewed papers：0

Type：Competition, symposium, etc. 

Type：Competition, symposium, etc. 

Type：Competition, symposium, etc. 

Type：Peer review 

Number of peer-reviewed articles in foreign language journals：8

Number of peer-reviewed articles in Japanese journals：0

Proceedings of International Conference Number of peer-reviewed papers：10

Proceedings of domestic conference Number of peer-reviewed papers：0

Type：Competition, symposium, etc. 

Type：Competition, symposium, etc. 

Type：Competition, symposium, etc. 

Type：Peer review 

Number of peer-reviewed articles in foreign language journals：5

Number of peer-reviewed articles in Japanese journals：0

Proceedings of International Conference Number of peer-reviewed papers：12

Proceedings of domestic conference Number of peer-reviewed papers：0

Type：Competition, symposium, etc. 

Type：Competition, symposium, etc. 

Type：Competition, symposium, etc. 

Type：Peer review 

Number of peer-reviewed articles in foreign language journals：0

Number of peer-reviewed articles in Japanese journals：0

Proceedings of International Conference Number of peer-reviewed papers：12

Proceedings of domestic conference Number of peer-reviewed papers：0

Type：Competition, symposium, etc. 

Number of participants：50

Type：Competition, symposium, etc. 

Type：Competition, symposium, etc. 

Number of participants：30

Type：Peer review 

Number of peer-reviewed articles in foreign language journals：2

Number of peer-reviewed articles in Japanese journals：0

Proceedings of International Conference Number of peer-reviewed papers：5

Proceedings of domestic conference Number of peer-reviewed papers：0

Introducing Learning Classifier Systems: Rules that Capture Complexity
Conference: GECCO 2018
(Tutorial in Top Conference)

Evolutionary Reinforcement Learning: General Models and Adaptation
Conference: GECCO 2018
(Tutorial in Top Conference)

創設者