Keiki Takadama

• Information Technology Center The University of Tokyo

Unlike the conventional swarm or evolutionary optimizations that are generally assumed the “pre-defined” bounded search space, this paper addresses the optimization for the “unbounded” search space. For this purpose, this paper proposes novelty-based multi-objectivization with local and rough area search (NM-LRS), which adds the novelty criterion i...

Rule representations significantly influence the search capabilities and decision boundaries within the search space of Learning Classifier Systems (LCSs), a family of rule-based machine learning systems that evolve interpretable models through evolutionary processes. However, it is very difficult to choose an appropriate rule representation for ea...

This paper focuses on “human-compatible AI” which aligns with human values and remains under human control to pre-vent unintended and harmful consequences, and discusses it to develop human-compatible AI for well-being. For this is-sue, this paper proposes the human-compatible AI for a sleep as one of the human-compatible AI for well-being, which i...

At the AAAI Spring Symposium 2025, we explored the challenges of integrating Human-Compatible AI and AI-Powered Science to enhance social and individual well-being. Our discussion was guided by two perspectives. Individual Impact of AI on Well-being: This perspective examines how AI influences personal autonomy, mental health, and emotional fulfill...

To introduce the concept of the “constraint tolerance” (i.e., a feasibility of solutions) in the flight scheduling problem, this paper proposes the optimization method that can find the feasible flight schedules by optimizing the original objective function while maximizing the constraint tolerance as much as possible. The proposed method further i...

Rule representations significantly influence the search capabilities and decision boundaries within the search space of Learning Classifier Systems (LCSs). However, it is very difficult to choose an appropriate rule representation for each problem. Additionally, some problems benefit from using different representations for different subspaces with...

The design of automated landscape-aware techniques requires low-cost features that characterize the structure of the target optimization problem. This paper approximates network-based landscape models of multi-objective optimization problems, which were constructed by full search space enumeration in previous studies. Specifically, we propose a sam...

This paper explores an answer to the question of “what is a correct output by generative AI from the viewpoint of well-being?” and discusses an effectiveness of taking account of a biological rhythm for this issue. Concretely, this paper focuses on an estimation of the REM sleep stage as one of sleep stages, and compared its estimations based on ra...

At the AAAI Spring Symposium 2024, we explore the important challenges facing Generative Artificial Intelligence (GenAI) concerning both social structures and individual welfare. Our discussion revolves around two perspectives. Individual Impact of GenAI on Well-being: This perspective focuses on the design of AI systems with keen consideration for...

This paper proposes the method by physiological knowledge to improve the estimation performance of the NREM3 sleep based on the waist-attached accelerometer. Specifically, this paper proposes the hybrid method that combines the method based on body movement counts and the method based on biological rhythms of sleep. Through the human subject experi...

As a first step towards realizing an AI sleep counselor capable of generating personalized advice, this paper proposes a method for monitoring daily sleep conditions with a mattress sensor. To improve the accuracy of sleep stage estimation and to get accurate sleep structure, this paper introduced sleep domain knowledge to machine learning for impr...

Inverse reinforcement learning (IRL) estimates a reward function for an agent to behave along with expert data, e.g., as human operation data. However, expert data usually have redundant parts, which decrease the agent’s performance. This study extends the IRL to sub-optimal action data, including lack and detour. The proposed method searches for n...

Since deep Q-networks and AlphaGO by Google DeepMind were proposed, not only a reinforcement learning integrated with deep learning but also its applications have attracted much attention on. ChatGPT, which learns by Proximal Policy Optimization as one of reinforcement learning mechanisms, is an excellent example. With this background, we propose t...

To increase an accuracy of the sleep stage estimation without connecting any devices/electrodes to the body, this paper proposes the updating method for its estimation according to an ultradian rhythm as one of the biological rhythms of humans. In the proposed method, the prediction probability of the sleep stage is updated by the Widrow–Hoff learn...

This paper focuses on Sleep Apnea Syndrome (SAS) and proposes the novel eXplainable AI (XAI) method that extracts characteristics of SAS by comparing the datasets of the SAS patients and the non-SAS subjects. For this issue, this paper (i) employs “two” Random Forests (RFs) to respectively learn the models for the SAS patients and the non-SAS subje...

Multi-agent Reinforcement Learning is required to adapt to the dynamic of the environment by transferring the learning outcomes in the case of the non-communicative and dynamic environment. Profit minimizing reinforcement learning with the oblivion of memory (PMRL-OM) enables agents to learn a co-operative policy using learning dynamics instead of...

People have been increasingly interested in sleep, and the number of services that provide simple sleep status monitoring is increasing. In order for users to continue to use these services, it is necessary to ensure that both users and services grow together. To achieve this, it is important to provide sleep state that is acceptable for people. To...

The cortical learning algorithm (CLA) is a time series prediction algorithm. Memory elements called columns and cells discretely represent data with their state combinations, whereas linking elements called synapses change their state combinations. For tasks requiring to take actions, the action-prediction CLA (ACLA) has an advantage to complement...

In this paper, a burst-based multilayered cortical learning algorithm (BM-CLA) for forecasting trend-changing time-series data is proposed. CLA predicts time-series data while adjusting synapse relationships online. However, the forecast accuracy of the conventional CLA deteriorates with trend-changing time-series data, in which several time-series...

This paper investigated the impacts of multi-objectivization on solving combinatorial single-objective NK-landscape problems with multiple funnel structures. Multi-objectivization re-formulates a single-objective target problem into a multi-objective problem with a helper problem to suppress the difficulty of the target problem. This paper analyzed...

This paper proposes the adaptive synapse adjustment and the adaptive decoding in the action-prediction cortical learning algorithm (ACLA) for an uncertain environment with probabilistically missing multiple input state values. The increase in the number of missing state values negatively affects the action prediction representation, and it is empha...

For an efficient upconvert of the Pareto front resolution by utilizing a known candidate solution set, this paper proposed an algorithm that built the Pareto front and the Pareto set estimation models and repeated to sample a solution from them, evaluate it, and updated the estimation models with it. Conventional supervised multi-objective optimiza...

This work introduces the following concepts of directional and estimated directional Pareto front to encourage multi-objective decision making, especially when the Pareto front exists in limited regions in the objective space. The general output of multi-objective optimization is a set of non-dominated solutions to approximate the Pareto front. Whe...

Air traffic flies from Asia to North America via the North Pacific (NOPAC) route system in oceanic airspace. The cruise altitudes of NOPAC routes are assigned on a first-come first-served basis. As a result, when an overflight and Japan departure flight compete for a cruise altitude, the former tends to receive its requested altitude, while the lat...

This paper focuses on the REM sleep estimation with bio-vibration data acquired from mattress sensor, and proposes its "correction" method based on Time-Series Confidence (TSC) of the REM sleep prediction calculated by Random Forest (RF) as one of the Machine Learnings (MLs). Unlike the conventional MLs that classify whether the REM sleep or not as...

It is important to detect daily Alzheimer dementia (AD) possibility using unconstrained mattress sensors because dementia takes time before subjective symptoms appear and the main treatment is to slow the rate of progression. Forcusing on circadian rhythm disorder which tend to occur with AD, this paper analyzes the features of unstable circadian r...

Social media is popular for us to share some news; however, it is easy for us to receive much fake news and believe them because of its simplicity. A new model is proposed for simulating a cyber-physical system preventing fake news with humans and agents by expanding the opinion sharing model (OSM), and this paper proposes a decision-supporting sys...

This paper focuses on the covering mechanism which generates a new if-then rule when the input data does not match the rules in the XCS Classifier System (XCS), a rule-based machine learning system, and discusses how the new rule should be generated from the viewpoint of “inheritance” and “expansion” of the generalization degree of the nearest neig...

This paper proposes the clustering-based optimization method for landing sequence of aircraft, which partitions all the aircraft into several clusters and optimizes the schedules of these parted aircraft in parallel. We conducted the computer simulation of the Charles de Gaulle Airport in France and revealed that (1) our proposed method obtains the...

The mission of this chapter is to formalize multi-objective reinforcement learning (MORL) problems where there are multiple conflicting objectives with unknown weights. The objective is to collect all Pareto optimal policies in order to adapt them for use in a learner's situation. However, it takes huge learning costs in previous methods, so this c...

To improve the accuracy to prevent from sharing incorrect opinion, this paper proposes a method which can share correct opinions based on majority decision for multi-opinion, named Gradient Descent Weight Tuning (GDWT). In the experiment, this paper compares GDWT with AAT and Self-information Weight Tuning (SWT) which weights the opinion from the a...

This paper proposes the novel Alzheimer dementia (AD) detection method based on unstable circadian rhythm of heartrate acquired from mattress sensor. Concretely, the pro-posed method, UCRADD (Unstable Circadian Rhythm based Alzheimer Dementia Detection), estimates the circadian rhythm of heartrate by calculating the regression of the trigonometric...

This paper proposes the novel Sleep Apnea Syndrome (SAS) detection method based on the frequency analysis of the overnight bio-vibration data acquired from mattress sensor. Concretely, this paper designs the index called Degree of Convexity of the Logarithmic Spectrum (DCLS), which quantifies the degree of convexity by computing the difference betw...

This paper establishes directionality reinforcement learning (DRL) technique to propose the complete decentralized multi-agent reinforcement learning method which can achieve cooperation based on each agent's learning: no communication and no observation. Concretely, DRL adds the direction "agents have to learn to reach the farthest goal among reac...

This paper focuses on the “early stage” of the online communication to investigate what kind of factors that contribute to forming a consensus among people who have their own way of thinking. For this purpose, this paper employs Barnga as the cross-cultural game where the players should select the winner according to their own rules, and analyzes o...

Reinforcement learning (RL) enables an agent to learn from trial-and-error experiences toward achieving long-term goals; automated planning aims to compute plans for accomplishing tasks using action knowledge. Despite their shared goal of completing complex tasks, the development of RL and automated planning has been largely isolated due to their d...

This paper proposes multi-factorial distance minimization problems for benchmarking of multi-factorial optimization. The multi-factorial optimization simultaneously searches for optimal solutions of multiple objective functions in the common variable space and is recently a popular issue regarding evolutionary optimization. The conventional multi-f...

This paper proposes the duplex route generation method to evolve the bus route network which is robust to environmental changes and aims at investigating its effectiveness through the experiments. In this study, the “duplex route” corresponds to the alternative route and it has the advantage of not requiring to modify the route network in the envir...

This work proposes a method to estimate the Pareto front even in areas without objective vectors in the objective space. For the Pareto front approximation, we use a set of non-dominated points, objective vectors, in the objective space. To finely approximate the Pareto front, we need to increase the number of objective vectors. It is worth to esti...

This paper explores the key-factors that can promote people to form a consensus remotely in such an Internet environment, and designs the agent according to the found key-factor for a better online communication. To address this issue, this paper focuses on “declaration of intent” and regards that people form a consensus when sharing one thought wi...

This work proposes a multi-factorial evolutionary algorithm encouraging crossovers among solutions with similar target objective functions and suppressing crossovers among solutions with dissimilar target objective functions. Evolutionary multi-factorial optimization simultaneously optimizes multiple objective functions with a single population, a...

This work proposes a double-layered cortical learning algorithm. The cortical learning algorithm is a time-series prediction methodology inspired from the human neuro-cortex. The human neuro-cortex has a multi-layer structure, while the conventional cortical learning algorithm has a single layer structure. This work introduces a double-layered stru...

This paper reports a relationship between emotional expressions and consensus building in virtual communication. Concretely, we focus on emotions before consensus-building. To investigate the relationship, we employ Barnga which is one of card games for experiments. In Barnga, players cannot use language, and they have not the same rule. In additio...

This paper proposes the information sharing algorithm for preventing propagation of wrong information in the agent-based network such as SNS, and aims at investigating the effectiveness of the proposed algorithm through the complex network such as a small world network. Towards practical applications, this paper extends the conventional opinion sha...

This paper proposes a goal selection method to operate agents get maximum reward values per time by noncommunicative learning. In particular, that method aims to enable agents to cooperate along to dynamism of reward values and goal locations. Adaptation against to these dynamisms can enable agents to learn cooperative actions along to changing tra...

This paper proposes a novel master–slave parallel evolutionary algorithm (EA) approach with different asynchrony and provides its detailed analyses on multi-objective optimization problems. We express the proposed EA with different asynchrony as a semi-asynchronous EA. A semi-asynchronous EA generates new solutions whenever evaluations of the prede...

This paper proposes a method of adaptation for a problem holds local optima with different properties around them. We employed neighborhood based method into Adaptive DE:JADE with a mechanism of local adaptation simultaneously at different local optima. Experimental results revealed (i) local search is effective to multiple properties problem aroun...

This paper extended PMRL as the non-communicative and theoretical method for two agents, and proposed PLA as the method to be able to force agents to learn cooperative behavior for any number of agents. In addition, this paper adds the theoretic explanation for PLA that all agents achieve all purposes without spending the largest times. Concretely...

This chapter describes solving multi-objective reinforcement learning (MORL) problems where there are multiple conflicting objectives with unknown weights. Previous model-free MORL methods take large number of calculations to collect a Pareto optimal set for each V/Q-value vector. In contrast, model-based MORL can reduce such a calculation cost tha...

This paper proposes a multi-agent reinforcement learning method without communication toward dynamic environments, called profit minimizing reinforcement learning with oblivion of memory (PMRL-OM). PMRL-OM is extended from PMRL and defines a memory range that only utilizes the valuable information from the environment. Since agents do not require i...

For solving multi-objective optimization problems with evolutionary algorithms, the decomposing the Pareto front by using a set of weight vectors is a promising approach. Although an appropriate distribution of weight vectors depends on the Pareto front shape, the uniformly distributed weight vector set is generally employed since the shape is unkn...

To briefly represent a dataset, it is crucial to find common attributes among the data. Extended learning classifier system (XCS) finds common attributes of multiple data and acquires generalized rules that match multiple data. In real-world problems, it may be challenging to find common attributes due to noise in the data and the inability of XCS...

This paper focuses on Omoiyari in Japanese as consideration/thoughtfulness for others in order to promote people to obtain a consensus among them especially in Internet society where is difficult to reach a consensus due to the limited communication/interaction, and aims at exploring the preliminary agent design that can promote people to obtain a...

This paper describes solving multi-objective reinforcement learning problems where there are multiple conflicting objectives with unknown weights. Reinforcement learning (RL) is a popular algorithm for automatically solving sequential decision problems and most of them are focused on single-objective settings to decide a single solution. In multi-o...

In data mining, it is important to clarify how effective the acquired rules are and which elements are affected by rule evaluation. Extended learning classifier system (XCS) reveals factors that affect the classifier (rule) evaluation by generalizing the multiple classifiers that acquire the same reward (evaluation value) into a generalized classif...

In this paper, we proposed Bat Algorithm extending with Dynamic Niche Radius (DNRBA) which enables solutions to locate multiple local and global optima for solving multimodal optimization problems. This proposed algorithm is designed Bat Algorithm (BA) dealing with the exploration and the exploitation search with Niche Radius which is calculated by...

This paper focuses on the artificial bee colony (ABC) algorithm as one of swarm optimization methods and proposes ABC-alis (ABC algorithm based on adaptive local information sharing) by improving the ABC algorithm for dynamic optimization problems (DOPs). ABC-alis is applied to various types of dynamic changes embedded in DOPs to verify its trackin...

In mechatronics and robotics, one of the important issues is to design human interface. There are two issues on interaction design research. One is the way to education and training to adapt humans for operating the robots or interaction systems. Another one is the way to make interaction design adaptable for humans. This chapter research at the la...

In Artificial Intelligence and Robotics, one of the important issues is to design Human interface. There are two issues, one is the machine-centered interaction design to adapt humans for operating the robots or systems. Another one is the human-centered interaction design to make it adaptable for humans. This research aims at latter issue. This pa...

Toward learning cooperative behavior for any number of agents, this paper proposes a multi-agent reinforcement learning method without communication, called PMRL-based Learning for Any number of Agents (PLAA). PLAA prevents from agents reaching the purpose for spending too many times, and to promote the local multi-agent cooperation without communi...

In recent years, massive earthquakes struck Japan, causing large-scale disasters such as the great Hanshin-Awaji earthquake in 1997, the Niigata Prefecture Chuetsu earthquake in 2004, the great east Japan earthquake in 2011, and the Kumamoto earthquake in 2016. In the all disasters above, logistics system for relief supplies collapsed and it was re...