[HOME],
[START]
BioFinder Kategorien Suche
Künstliches Leben < Biomathematik
< Biologie
Links in dieser Kategorie:
- GWAL 3rd German Workshop on Artificial Life
Unversity of Bochum, Haus Bommerholz, 17.-18. September 1998
- A Multi-agent Approach to Fuzzy Modelling
This project investigates the techniques for structural and parametric identification in an attempt to represent and reason about vague knowledge with fuzzy production rules. It aims to create an automated system for generating fuzzy rules from numerical data with minimal human intervention. Current investigation follows a multi-agent approach, applying a clustering algorithm to form the initial membership functions of the fuzzy values, developing a rule induction mechanism to construct the initial rule base, and adapting an evolutionary computation method to perform the fine-tuning of the resulting membership functions and rules. To test the ideas, the multi-agent system will be utilised to solve realistic problems in engineering and financial domains.
- ALife Garden
- ALife Online
- ALife Playroom
- AI Artificial Intelligence
- ALife Bibliography Artificial Life Bibliography
- Artificial Painter
- artificial painter
artificial painter Artificial Life techniques can also be used to evolve aesthetic pictures to be used in artistic design. The evolution of pictures is based on the user's aesthetic evaluation of a number of pictures shown on the screen. The Artificial Painter model originates from an inspiration of bio-image techniques such as Computed Tomography (CT), Positron Emission Tomography (PET), and Single Neuron Records. In particular, in AP the techniques of single neuron records are applied on an Artificial Neural Network. The different activation of units in different neural networks gives rise to different, colourful pictures shown to the user on the computer screen. The user can select a number of these pictures to reproduce, and thereby guide the evolution process toward the individual estimate of fitness.
- Avida Group of CalTech
- Virtual Barley Barley Modelling and Simulation
On this site, a computergraphic model is presented that - based on the vlab - is visualizing barley spike (Hordeum vulgare L.) phenotypes using genotype information. Also,a set of sensitive maps containing all morphological mutants is provided.
- Bionic SONAR Head
The work done on the bionic SONAR head has a dual purpose: 1. to provide a platform from which to investigate the real-time information processing tasks carried out by biological systems -- i.e., echolocating bats; 2. to engender artificial navigation systems with some of the sophisticated performance and robustness of those biological systems.
- C.O.P.P. Software
- Causal Explanation for Model-based Diagnoses of Dynamic Systems
Process supervision becomes increasingly important as a major factor affecting the efficiency of the manufacturing industries. An important problem in process supervision is the diagnosis of faults affecting the normal operation of the plant. Model-based diagnosis (MBD) allows the automatic location of faults by reasoning about explicit models of the plant under consideration. This approach offers many advantages over the alternative of relying on a rule base of anticipated or previously encountered malfunctions. Although much research has been done on the development of the diagnostic techniques, the area of explanation in MBD has been largely unexplored, especially in diagnosing dynamic systems containing feedback loops. This research aims to develop an explanation mechanism which will reveal the causal relationships between the diagnostic findings, returned by an MBD system working in a dynamic environment, and the observed behaviour of the plant. The focus of the work is to ensure that the generated explanations will achieve two fundamental design requirements: a) offering a meaningful causal interpretation of the found fault, explaining not just what fault model reflects the observed faulty behaviour of the plant, but also how and why, and b) presenting information in a format that is comprehensible to the human user. This approach to self-explanatory fault diagnosis has the potential of helping increase the effectiveness and efficiency of process supervision in industrial applications.
- Control of locomotion
Walking: This work is looking at the possibility of controlling a legged robot with a structure based on that used in vertebrates, i.e. Central Pattern Generators - these are neural networks which generate the basic rhythmical motor pattern for walking adapted to the simple high level control signals they receive from the brain and the sensory feedback they receive from the body. No specific animal or CPG is being directly modelled, the aim is rather to try to apply ideas gleaned from neurophysiological research into these structures to the problem of controlling legged robots.
- Cricket phonotaxis
Cricket phonotaxis The aim of this work is to use robots to model specific biological sensorimotor control systems. By taking a robotic approach (how can I get a machine to behave this wayNULL) to a well-explored biological problem (what are the known characteristics and underlying systems for this behaviourNULL) we can draw on the strengths of both fields in attempting to understand how perceptual mechanisms work.
- evolution of parallel self-replicating programs
An environment is being developed to facilitate the study of patterns of evolution in a system of self-replicating entities which are competing with each other for resources. In this environment, computer programs are the self-replicators, and large numbers of them compete with each other for the memory and CPU time required to make copies of themselves. The programs are also subject to mutation, so that, over time, mutants which are better at making copies of themselves become more numerous in the population of programs, and a process of evolution is observed.
- Evolution von Pflanzen (virtuell)
- evolutionary robotics
evolutionary robotics Are robots technical devices that have to be developed and controlled by a human engineer, or could robots also develop and control themselves autonomouslyNULL Traditional robotics that uses Artificial Intelligence planning techniques to program robot behaviors works toward the first, while the Autonomous Robotics approach suggest that the second is a possibility. The robots built according to this approach should be able to adapt to both uncertain and incomplete information in constantly changing environments.
- Generating Explanations of Industrial Processes Using Multiple Models
This project aims towards building an intelligent system with the ability to provide informative explanations for industrial processes, based on a multiple models representation of the supporting domain knowledge. The work pursues a further expansion of the current approaches to model-based reasoning, with respect to the task of automatic model formulation from existing knowledge. The system under development attempts to construct domain models by combining model fragments, i.e. models for different components and/or processes at various detail levels. The initial fragments of a model under construction are selected according to the requirements of the users (e.g. plant operators), and the system's assessment of their expertise. The selection of the other required fragments is completed by exploring a Bayesian belief network. The resulting model is then used as the source of the necessary knowledge for the generation of explanations.
- Generating Fuzzy Classification Rules from Crisp Examples
Rule induction from historical data provides an important key to the solution of the knowledge acquisition bottleneck in the development of knowledge-based systems. Numerous studies have been made on generating rules from example data, in research fields ranging from symbolic machine learning to non-symbolic connectionist computing. However, algorithms developed in machine learning are mainly for discrete domains, whilst neural network-based approaches work by making strong assumptions such as rules involving only one premise attribute or a small number of attributes which are pre-selected by a statistical method. Alternative approaches are necessary, in order to produce rules capable of being applied to situations where data are real-valued, multi-dimensional and uncertain. This project sets to investigate a recently proposed technique for rule induction which aims at extracting linguistically expressed rules from real-valued examples. The original proposal was largely tailored to the application of neural network-based classifiers, but appears to be independent of the actual form of the network mapping. The present investigation is, therefore, to explore the potential of generalising this technique to solving real problems concerning knowledge discovery in a given domain, e.g. in the financial service sector for consumer modelling. Within this project, the original proposal for rule induction is formalised and re-engineered in a context-free environment, and its performance is evaluated with respect to typical classification problems. Also, a mechanism for further condensing the induced rules is devised, by integrating the re-engineered version with a standard machine learning algorithm. This work is being carried out within the framework of a general purpose knowledge-based system, which allows for approximate reasoning to be performed when given a knowledge base consisting of the induced, and condensed, rules for a problem domain.
- Intelligent Strategist for Qualitative Model-based Diagnosis
Model-based diagnostic systems use the discrepancies between the observed and predicted behaviours, of the physical system under diagnosis, to postulate fault candidates. However, the fault space for a diagnostic system to search is usually large and so a strategist is required to control the search. An aim of this project is to develop such a strategist that can guide the diagnostic systems in order to help them find the correct fault models that minimise the discrepancies as quickly and efficiently as possible. The strategist is designed to be capable of `learning' how to guide a diagnostic process from past experience, based on a combination of self-organising fuzzy logic control and fuzzy system identification techniques. Another major aspect of the project is the use of Markov chains for updating certainty beliefs associated with different models of a physical system. This not only provides a concrete basis for the strategist to work on, but also actively extends the ability of a diagnostic system by providing a means of searching along the uncertainty dimension.
- Lifepage
- LogicAL Logic, Philosophy, and Artificial Life Resources
- MIT Press Journal
- Modelling Organisations and their Operations
This project is currently at its initial phase. The present work is focused on investigating the appropriateness of using qualitative and approximate modelling techniques to represent different organisations behaviours from the perspective of applied economics. The work takes as input lessons learned from modelling engineering systems with qualitative and/or approximate approaches. As suitable techniques are identified, they will be combined and adapted to better suit the needs of describing organisations behaviours that are of abstract, uncertain and dynamic nature.
- neuroethological robotics
neuroethological robotics Where Evolutionary Robotics helps us to develop control systems automatically, Neuroethological Robotics can be used to verify the capabilities of known (or pre-defined) control systems. Based on neurophysiological or behavioral evidence, biologists and neuroethologists pose different hypotheses about different animals' control systems. In this context, Neuroethological Robotics provides an empirical field where testing such hypotheses by implementing them on real robots is made possible.
- Pea Experiment
- Putting RL onto robots
Recently a lot of work has been done trying to put RL algorithms onto real robots and there have been a number of successful implementations to date (DorigoColombetti93, Kaelbling90a, MaesBrooks90, MahadevanConnell90, Mataric94b, Nehmzow92) . There are, however, a number of difficulties associated with RL methods per se and these are especially pertinent to the problem of using them with real robots. In short RL makes assumptions which do not apply to real world tasks (cf. Mataric94b ). First RL assumes that the environment as perceived by the agent is a Markov Decision Process MDP. Informally this means that the agent only need know the current state of the process in order to predict its future behaviour. In principle any process can be represented as an MDP because an abitrary amount of information about the history of the process can be included in the description of the current state, eg. we need to know the velocity and acceleration of a ball in order to be able to calculate its trajectory. If the agent does not have sufficient information to predict the future process behaviour (in the case of RL this means it does not have sufficient information to predict average return accurately.) then what Whitehead terms perceptual aliasing (cf. WhiteheadBallard90) occurs. This is when an agent cannot distinguish between two states which are significantly different with respect to their behaviour under the same policy. New algorithms have been designed to cope with this phenomena (cf. LinMitchell92, WhiteheadBallard90) but are not guaranteed to converge to the optimal policy under such conditions.
- Reinforcement Learning
Reinforcement Learning (RL) is a trial and error approach to learning that has recently become popular with roboticists. This is despite the fact that RL methods are very slow, and scale badly with the size of the state and action spaces, thus making them difficult to put onto real robots. Reinforcement learning is a trial and error approach to learning in which an agent operating in an environment learns how to achieve a task in that environment. The agent learns by adjusting its policy (The mapping from environment states to agent actions) on the basis of positive (or negative) feedback --- termed reinforcement. This feedback takes the form of a scalar value generated by the agent each time step, high and low values corresponding to rewards and punishments respectively. The mapping from environment states and agent actions to reinforcement values is termed the reinforcement function. The agent converges to the behaviour maximising reinforcement (the optimal policy). In theory an appropriate reinforcement function (ie. one in which the policy maximising reinforcement corresponds to the behaviour which the designer considers optimal) exists for all tasks, although finding such a function is typically hard (see Barto90b).
- GRAL Research Group on Artificial Life
- ALIFE VI Sixth International Conference on Artificial Life
ALIFE VI will be held in June of 1998 on the campus of the University of California
- SC & AL Soft Computing & Artificial Life
- speciation via habitat specialization
speciation via habitat specialization A long term goal in Artificial Life is to evolve speciation in Artificial Life models. An approach toward this might be to evolve assortative mating, since this is believed to lead to speciation. For instance, sympatric speciation may happen via habitat specialization. Our work on speciation in Artificial Life simulations has been to evolve habitat specialization in populations of neural networks. The emergence of specialization can be due to individual energy extracting abilities, and simulations show that the energy extracting mechanism, the sensory apparatus, and the behavior of organisms may co-evolve and be co-adapted. Populations of organisms have been shown to be pre-adapted to changing environments where the preferred food disappears, and an analysis of the activation of neural network hidden units show that a new food preference can be an ex-aptation, i.e. a new adaptation based on a structure which has previously emerged for adaptively neutral reasons. Further, under social conditions in shared environments it has been found that competition can act to provide population diversification in populations of organisms with individual energy extracting abilities.
- The Game of Life
- The Halperin net
Bionic SONAR Head
The Halperin net There has been a stimulating dialogue between automous robot researchers and ethologists (biologists studying animal behaviour). Our group has taken a particular interest in a model called the Halperin net.
- The Temple of ALife
- Uncertain Constraint Satisfaction Problems in Model-based Systems
Existing techniques for solving constraint satisfaction problems (CSPs) are largely concerned with quantities that can be represented by nominal symbols or crisp real intervals, and the constraints involved are normally well-defined over such quantities. In addition, the problem itself is typically assumed to be of a static nature. Recently, work has been done to address these shortcomings of classical constraint satisfaction in the form of two separate extensions known as flexible and dynamic CSP. Little work has been done to combine these two approaches in order to bring to bear the benefits of both in solving more complex problems. The aim of this project is, therefore, to develop a dynamic and flexible solution technique capable of supporting uncertain values and constraints, especially with fuzzy intervals and relations, and which can also deal with changes to the structure of the problem over time. The resulting mechanism will be embedded in a system designed to perform planning tasks in an uncertain environment subject to continual change.
- Virtual ALife Library
- Zoo Land
© 2000 by Kurt Stüber