Current Talks
|
Date & Location
|
Topic & Presenter(s)
|
04/25/08
1:00PM
AVW 3258
|
TBD
Grecia Lapizco-Encinas
|
| Abstract
TBD
03/07/08
12:00PM
AVW 4185
|
Using Swarm Intelligence to Grow Artificial Neural Networks
Charles E. Martin
|
| Abstract
The process by which a highly complex biological neural network develops from a
small number of initial cells is governed by two dependent processes; both the
expression of genetic information contained within each developing neuron and
the neurons' interactions with their environment are responsible for the network
that emerges. Models that describe neural growth employ various degrees of
abstraction. While some attempt to describe the underlying biological mechanisms
in great detail, others focus more on capturing the correct behaviors and less
on the biological realism of the mechanisms generating them. In this talk I will
discuss the benefits of using an agent based model of neural network growth and
demonstrate the effectiveness of a software based simulator that implements the
model. Particular attention will be focused on the model's novel incorporation
of L-system principles and swarm intelligence to provide highly controllable
growth of large networks. The model represents axon growth and neuron migration
explicitly in a continuous three-dimensional space. The representation of
growing axons and migrating neurons as continuous events provides an important
degree of realism to the neurogenesis, while implementing the L-system and swarm
intelligence mechanisms provides significant improvements to prediction and
controllability. A proposed extension to the simulator for optimizing neural
network topologies will also be discussed.
02/29/08
12:00PM
AVW 4185
|
Multidirectional Symmetry in Self-Organizing Multi-Maps
Jared Sylvester
|
| Abstract
The sensory cortices of numerous species are arranged into adjacent, symmetrically oriented topographic maps of the corresponding sensory surface. However, the mechanism responsible for this process is
unknown. I will present a computational model which suggests such symmetries can arise purely as a result of afferent activity. The model in question is a Kohonen Self-Organizing Map which has been
modified to allow multiple winning nodes for each input stimulus. This technique allows multiple topographic representations of the sensory surface to emerge, with neighboring maps orienting themselves
into two dimensional grids of individual maps. Four different types of symmetrical orientation result, including one which was previously unobserved in which two maps do not form a firm boundary and
instead have merged together, while still maintaining proper topographical relationships. The effect of the symmetry relationships on map formation, as well as the effects of non-uniform sensory
stimulation are also considered.
11/09/07
1:00PM
AVW 3258
|
An Introduction to Echo State Neural Networks
Charles E. Martin
|
| Abstract
Though recurrent neural networks have been demonstrated to be useful on many computational tasks, particularly those for which there is a time dependence amongst the inputs to the network, the difficulty
in effectively and efficiently training them has proven to be a formidable barrier. Multiple avenues have been taken in response to this challenge. For example, the well known Back-Propagation method for
feedfoward networks has been extended to recurrent networks and specific network topologies have been designed to implement models from the field of signal processing. However, significant difficulties
still persist. Echo state networks represent a class of large, non-modular recurrent neural networks. They have emerged as a means of harnessing the computational power of recurrent neural networks,
without the need to train recurrent connections. Though the formal echo state network model is relatively new, its usefulness has already been demonstrated on a variety of tasks involving system
identification and time-series prediction. This talk will cover the basics of echo state networks, including the architectures, network properties, training methods and applications.
10/12/07
1:00PM
AVW 3258
|
Improving Neural Network Encoding by Regularized Training
Thuan Huynh
|
| Abstract
Error backpropagation is the most widely used supervised learning method for neural networks and
has achieved success in many classification, regression and prediction applications. The network
learns a mapping between the input and output units, while the hidden units and the weights
between them and other units contain the network's internal representation of the input. This
work addresses the problem that the neural network's internal representations learned by error
backpropagation are very difficult for people to interpret. Our approach is to add a
regularization term to the error function used in training to create a sparser encoding at the
hidden layer in neural networks. This term helps to increase the sum
of squared/inverted squared Euclidean distances between hidden activation vectors. The method was
applied to public datasets from the UCI machine learning repository. We show that the final
encodings are easier to distinguish and provide a good basis for learning symbolic rules from the
network. We also applied it to the phoneme sequence recognition problem and got some promising
initial results.
09/28/07
11:30AM
AVW 3258
|
An Oscillatory Hebbian Network Model of Working Memory
Ransom Winder
|
| Abstract
We developed an oscillatory Hebbian network featuring a rapid decay of connection strengths during training and studied this as a computational model of working memory
capacity that performs a simulated running memory span task. This decay allowed the network to have a dependency on the order of input and preserve recent items active for
later recall. This model demonstrates recall performance similar to humans performing the same task, with a prominent recency effect and a capacity limit of approximately
three items. The results of the model are compared with data from two behavioral experiments that explored working memory capacity limits with varying task demands.
Behavioral data support that if task demands require attention to be spread too thin to keep stimuli available in memory for later recall then capacity limits suffer, a
result also observable in the model, which requires decay to curtail the inherent interference. These findings are important for extending research into oscillatory networks,
understanding the mechanisms underlying short term memory capacity, and also to memory researchers interested in the role of attention in capacity limitations.
08/10/07
1:00PM
AVW 3165
|
Artificial Evolution of Arbitrary Self-Replicating Cellular Automata
Edward Z. Pan
|
| Abstract
Since John von Neumann's seminal work on developing cellular automata models of self-replication, there have
been numerous computational studies that have sought to create self-replicating structures or "machines".
Cellular automata (CA) has been the most widely used method in these studies, with manual designs yielding a
number of specific self-replicating structures. However, it has been found to be very difficult, in general, to
design local state-transition rules that, when they operate concurrently in each cell of the cellular space,
produce a desired global behavior such as self-replication. This has greatly limited the number of different
self-replicating structures designed and studied to date. In this dissertation, I explore the feasibility of overcoming this difficulty by using genetic programming (GP)
to evolve novel CA self-replication models. I first formulate an approach to representing structures and rules
in cellular automata spaces that is amenable to manipulation by the genetic operations used in GP. Then, using
this representation, I demonstrate that it is possible to create a "replicator factory" that provides an
unprecedented ability to automatically generate whole families of self-replicating structures and that allows
one to systematically investigate the properties of replicating structures as one varies the initial
configuration, its size, shape, symmetry, and allowable states. This approach is then extended to incorporate
multi-objective fitness criteria, resulting in production of diversified replicators. For example, this allows
generation of target structures whose complexity greatly exceeds that of the seed structure itself. Finally,
the extended multi-objective replicator factory is further generalized into a structure/rule co-evolution
model, such that replicators with unspecified seed structures can also be concurrently evolved, resulting in
different structure/rule combinations and having the capability of not only replicating but also carrying out a
secondary pre-specified task with different strategies. I conclude that GP provides a powerful method for
creating CA models of self-replication.
07/20/07
1:00PM
AVW 3165
|
The Influence of Collective Working Memory Strategies on Agent Teams
Ransom Winder
|
| Abstract
Past self-organizing models of collectively moving "particles" (simulated bird flocks, fish schools, etc.) typically have been
based on purely reflexive agents that have no significant memory of past movements or environmental obstacles. These agent collectives usually operate in
abstract environments, but as these domains take on a greater realism, the collective requires behaviors use not only presently observed stimuli but also
remembered information. It is hypothesized that the addition of a limited working memory of the environment, distributed among the collective's individuals
can improve efficiency in performing tasks. This is first approached in a more traditional particle system in an abstract environment. Then it is explored
for a single agent, and finally a team of agents, operating in a simulated 3-dimensional environment of greater realism. In the abstract environment, a
limited distributed working memory produced a significant improvement in travel between locations, in some cases improving performance over time, while in
others surprisingly achieving an immediate benefit from the influence of memory. When strategies for accumulating and manipulating memory were subsequently
explored for a more realistic single agent in the 3-dimensional environment, if the agent kept a local or a cumulative working memory, its performance
improved on different tasks, both when navigating nearby obstacles and, in the case of cumulative memory, when covering previously traversed terrain. When
investigating a team of these agents engaged in a pursuit scenario, it was determined that a communicating and coordinating team still benefited from a
working memory of the environment distributed among the agents, even with limited memory capacity. This demonstrates that a limited distributed working
memory in a multi-agent system improves performance on tasks in domains of increasing complexity. This is true even though individual agents know only a
fraction of the collective's entire memory, using this partial memory and interactions with others in the team to perform tasks. These results may prove
useful in improving existing methodologies for control of collective movements for robotic teams, computer graphics, particle swarm optimization, and
computer games, and in interpreting future experimental research on group movements in biological populations.
03/09/07
12:00PM
AVW 3258
|
Antibody Selection, Danger Theory, and Recruitment in Artificial Immune Systems
Cole Trapnell
|
02/16/07
12:00PM
AVW 3258
|
Adapting Swarm Intelligence for the Self-Assembly of Prespecified Artificial Structures
Alexander Grushin
|
01/26/07
12:00PM
AVW 3258
|
Associative Networks for a Short Term Memory Model
Ransom Winder
| | | | | | | | |
|
|
Archive 2006
|
Date
|
Topic & Presenter(s)
|
| 12/01/06 12:00PM
|
TBA
Mike O'Hara
|
| 10/27/06
|
Extending Self-Organizing Particle Systems
for Basic Problem Solving
Alejandro Rodríguez
|
| 09/22/06
|
Neural Network Generation of Temporal Sequences from
Single Static Inputs using Varying Length Distal Target Sequences
Shaun Gittens
|
|
| Abstract
Training a learning agent to operate in an external environment whose mappings are largely unknown has been shown to be exceptionally difficult. Furthermore, granting such a learning agent the ability to produce some series of appropriate actions entirely from a single input stimulus has hardly been addressed. Various reinforcement learning techniques have often been utilized to handle such learning tasks, but convergence to optimal policies is not guaranteed for many of these methods. Traditional supervised learning methods hold more assurances of convergence to optimal policies than reinforcement learning, but these methods are not well suited for tasks such as these which require working in environments. This is because there is a tendency for desired actions in the output space of the learner not to be known or readily available for training by the proverbial ``teacher''. Rather, target outputs attainable through the environment are generally the only information the teacher can supply, leaving the learner to obtain the desired proximal sequential behavior required to yield them. Distal supervised learning techniques have been devised such that the strengths of traditional supervised learning methods may now be used for learning tasks in complex external environments. However, not much has been done to show how one can create a neural network that can be trained to produce sequences of actions from only a single input (no current state information provided) to yield desired target sequential behavior in an environment. Effective use of recurrent neural network learning strategies for distal settings is presented here for addressing these types of sequential behavior acquisition problems. One advantage seen in utilizing recurrent neural nets in problem domains such as this include the ability to learn to generate sequences of actions even in the absence of current or previous state information.
| 05/05/06
|
Supervised Learning Strategies for Sequence Generation in Neural
Networks
that Operate in Complex External Environments
Shaun Gittens
|
| 03/31/06
|
Using Aggregate Motion in Multi-Agent Teams to Solve Search and Transport Problems
Alejandro Rodriguez
|
|
| Abstract
The aggregate movement of animals has inspired computational models that have proven useful for controlling navigation
in teams of mobile agents. Additionally, strategies adopted from social insects have been employed in a multitude
of problems to perform distributed problem solving. In this work, we combine collective movements with general
problem solving capabilities to build multi-agent teams that perform search and collection tasks that require not only
navigational skills but also a coordinate team strategy. The results show that agents undertaking collective movements
are better able to self-coordinate and propagate information than those moving independently.
| 03/10/06
|
Simulating Multiple Word Processing Following the Wernicke-Lichtheim-Geschwind Model
Ransom Winder
|
| 02/24/06
| Automating the Design of Control Mechanisms for Self-Assembly Processes
Alexander Grushin
|
|
| Abstract
Recent years have witnessed an increasing level of attention being devoted to controlling the self-assembly of predefined target
structures via the use of local, agent-level behaviors. Given some target structure in a non-trivial environment, the design of
appropriate behaviors is a challenging problem, and presently, automated design procedures exist only for the simplest of environments,
and generally result in large sets of control rules. This talk presents an ongoing research effort to (a) extend such procedures to an
environment with a constrained, continuous motion, and (b) allow the generation of parsimonious rule sets. It is argued that
environmental constraints impose ordering constraints on the self-assembly process, where certain subsets of the target structure must
be assembled before others, and where it may be necessary to assemble (and subsequently disassemble) temporary components. These
constraints are abstractly captured in a simple mathematical model, and an ordering algorithm is presented, with provable (under
specific assumptions) properties such as correctness and completeness. Once an order is computed, rules are generated to enforce this
order in the course of self-assembly via the use of memory variables and local communication. The assembly and disassembly of
components is achieved by generating stigmergic rules, which are inspired by construction behavior amongst certain species of social
insects. The rule generation procedure is empirically tested on a set of six distinct target structures, with a generic set of movement
control mechanisms, and its performance is critically evaluated in terms of parsimony (the number of rules generated) and efficiency
(the average amount of time necessary to self-assemble the target structure, given the set of rules). These preliminary results are
discussed, with an outline of current and future research directions.
Articles
Chris Jones & Maja J. Mataric,
From Local to Global Behavior in Intelligent Self-assembly
In Proceedings of the IEEE International Conference on Robotics and Automation (ICRA'03) (Link to citeseer)
| | |
|
|
Archive 2005
|
Date
|
Topic & Presenter(s)
|
| 11/11/05
| The Neural Bases of Lateralization Effects in Visual Frequency Processing:
A Computational Modeling Investigation
Mary Howard
Abstract
Several psychophysical and electrophysiological experiments have provided evidence that the processing of frequency information in the
visual domain differs in the two hemispheres. The psychophysical results indicate that the processing of spatial frequencies is
lateralized, with a right hemisphere advantage for low and a left hemisphere advantage for high spatial frequencies. The
electrophysiological results suggest that temporal rather than spatial frequency is the key factor, with a right hemisphere advantage
for low and a left hemisphere advantage for high temporal frequencies. In this seminar, I present my research on the biological
underpinnings of these lateralization effects. I first discuss a neural network modeling experiment that demonstrates that differences
in the timing of development of the major visual pathways, together with asynchronous maturation of the hemispheres, could result in
the development of hard-wired asymmetry that lateralizes the processing of spatial frequencies. I then present a theoretical model that
explains how the hard-wired asymmetry, when combined with spatial attention, could account for the spatial frequency lateralization
effects observed in the psychophysical experiments. I go on to explain why the hard-wired asymmetry cannot account for the temporal
frequency lateralization effects observed in the electrophysiological experiments and discuss my hypothesis that these effects are the
product of gross anatomical asymmetries rather than actual processing differences. I present the results of a computational model of
the dipole-VEP wave relationship in support of this contention.
|
| 11/04/05
| Using Swarm Intelligence to Generate Artificial Neural Networks
Charles E. Martin
Abstract
In this project we will develop software that utilizes aspects of biologically inspired computing as a nontraditional approach to the
generation and evolution of artificial neural networks, (neural networks). Specifically, principles from swarm intelligence and genetic
algorithms or programming will be applied to generate and evolve a two dimensional neural network into a specified architecture. The
model, and the software based on it, will account for the geometry as well as the topology of the network. In addition to providing a
novel means of encoding neural networks and exploring different architectures, this project is intended to add to our understanding of
the relationship between the microscopic rules governing the interactions of a particle system consisting of locally interacting
autonomous agents and the resultant macroscopic behavior.
|
| 09/16/05
| Extending Swarm Intelligence to Abduction Problems
Grecia C. Lapizco-Encinas
Abstract
Swarm intelligence is a relatively new discipline, inspired by collecting moving animals, that deals with the study of collective
behavior in decentralized, self-organized systems. Algorithms inspired by these models have been proposed for simulation of biological
populations, robotic control, and discrete or continuous optimization problems. Recently, most research has focused on the application
and improvement of particularly successful methods, such as ant colony optimization or particle swarm optimization to real-world
problems This work proposes a different research direction. The central hypothesis in this proposal is to develop and study a new swarm
intelligence algorithmic approach for a task not usually tackled by this discipline: abductive problems. To verify this hypothesis the
proposed model will be first tested in a basic diagnosis task, the most representative application of abduction. This task will be
later extended and finally the model will be tested in another abduction application: a basic natural language processing task.
|
| 08/11/05
| Integrating Stigmergy and Coordinated Motion for Intelligent Self-Assembly: Design and Evolution
Alexander Grushin
Abstract
The emerging field of intelligent self-assembly addresses the fundamental problem of designing local behaviors that will cause
initially-unorganized components to form a desired structure. An effective solution to this problem would have considerable practical
significance, as it carries the promise of automating the construction of useful objects in environments that are dangerous or
inaccessible to human beings. However, the field is still far from reaching this potential, past studies presenting simulations with
either very idealized environments or very simple constructed objects. A central aim of this proposal is to narrow the gap that exists
between simulations and the demands of practical applications within the physical world. To this end, I will design and implement
control mechanisms that allow non-trivial 3D structures to self-assemble from different-sized blocks moving in a continuous environment
that simulates gravity and other physical constraints. These mechanisms integrate stigmergic pattern recognition, force-based movement,
and coordination via a very limited amount of memory; furthermore, they are entirely distributed, each block making independent
decisions based on local information. Factors influencing the efficiency of the self-assembly process (presence of noise, number of
components, etc.) and its robustness in the face of errors will be studied experimentally. Further research will be conducted on
automating the design of the block controllers through the use of evolutionary computation. I hypothesize that self-assembly can be
performed effectively without centralized intervention, and that an evolutionary approach is capable of designing new and improving
existing control mechanisms. Importantly, testing these hypotheses will yield a better understanding of the complex dynamics of both
self-assembly and evolutionary computation.
|
| 08/08/05
| Artificial Evolution of Arbitrary Self-Replicating Structures
Edward Z. Pan
Abstract
Our incomplete understanding of the complex mechanisms underlying reproduction in biological entities has led to numerous computational
studies over the last half century that have sought to create self-replicating structures or machines. Cellular automata (CA) has been
the most widely used method in these studies, with manual designs yielding a number of specific structures or configurations capable of
self-replication. It is extremely difficult, in general, to design local state-transition rules that, when they operate concurrently in
each cell of the cellular space, produce a desired global behavior such as self-replication. While previous attempts at automated
design of self-replicating structures using genetic algorithms have produced some interesting results, these studies were limited by
the enormous computational costs incurred. Here, I propose that genetic programming based on a suitable encoding of CA structures and
rules may provide a very powerful tool for discovering novel CA models of self-replicating systems and possibly other complex systems.
I describe some preliminary work using genetic programming methods to automatically discover CA rule sets that produce self-replication
of arbitrary given structures. The current experiments have produced larger, more rapidly replicating structures than past evolutionary
models while requiring only a small fraction of the computational time needed in past similar studies. Research methods are developed
and discussed for implementing a novel, general, artificial evolution model and quantitatively studying its variations and behavior,
for implementing multistage evolution toward debris-cleaning, for increasing the complexity of self-replication, and for the
co-evolution of both structure and rules.
|
| 06/24/05
| Genetic Programming and Games
Moshe Sipper, Ben-Gurion University
Abstract
Evolutionary algorithms are common nowadays, having been successfully applied to numerous problems from different domains, including
optimization, automatic programming, circuit design, machine learning, economics, immune systems, ecology, and population genetics, to
mention but a few. Applying the evolutionary methodology of genetic programming, wherein computer programs evolve, to three games ---
backgammon, chess (endgames), and robocode (tank-fight simulation) --- I show that evolved game players are able to hold their own, and
often win, against human or human-based competitors. Genetic programming is thus shown to be a prime candidate for evolving strategies
in games and other domains.
Brief Bio.
Moshe Sipper received the B.A. degree from the Technion -- Israel Institute of Technology, and the M.Sc. and Ph.D. degrees from Tel
Aviv University, all in computer science. He is currently an Associate Professor in the Department of Computer Science, Ben-Gurion
University, Israel. During the years 1995-2001 he was a Senior Researcher in the Swiss Federal Institute of Technology in Lausanne. Dr.
Sipper's major research interests include evolutionary computation, bio-inspired computing, and artificial life; minor interests
include cellular computing, cellular automata, and artificial self-replication, along with a smidgen of evolutionary robotics,
artificial neural networks, and fuzzy logic. Dr. Sipper has published over 110 scientific papers, and is the author of two books:
"Machine Nature: The Coming Age of Bio-Inspired Computing" and "Evolution of Parallel Cellular Machines: The Cellular Programming
Approach". He is an Associate Editor of the IEEE Transactions on Evolutionary Computation and an Editorial Board Member of Genetic
Programming and Evolvable Machines. Dr. Sipper is the recipient of the 1999 EPFL Latsis Prize.
|
| 05/12/05
| Diagnostic Problem Solving Using Swarm Intelligence
Grecia C Lapizco-Encinas
Evolutionary Discovery of Arbitrary Self-Replicating Structures
Edward Z. Pan
Using Aggregate Motion in Multi-Agent Teams to Solve Search and Transport Problems
Alejandro Rodriguez
|
| 05/05/05
| Modular Neural Network Models of Functional Brain Imaging
Ransom Winder
|
| 04/28/05
| A Cellular Automata Model of Network Markets
James A. Reggia
|
| 04/22/05
| Computational Maps in the Visual Cortex
Risto Miikkulainen, University of Texas
Abstract
How can a system as complex as the human visual system be constructed? How can it be specified genetically, still allowing it to adapt
to the environment? How can it perform complicated functions such as recognizing faces and identifying coherent objects immediately and
automatically? While these questions have been open for quite some time, and much experimental work remains to be done to answer them
conclusively, computational models have recently become powerful enough to suggests specific, computational answers: The cortical
structures are constructed through input-driven self-organization, the self-organization is driven both by external visual inputs and
by genetically determined internal inputs, and perceptual grouping takes place automatically through synchronization of neuronal
activity, mediated by self-organized lateral connections. In this talk, I will describe a unified computational map model, LISSOM,
built on these principles. Simulated experiments with LISSOM demonstrate how a wide variety of phenomena follow from them, including
columnar map organization and patchy connectivity, recovery from retinal and cortical injury, psychophysical phenomena such as tilt
aftereffects and contour integration, and newborn preference for faces. The model is used to gain a precise computational understanding
of existing data, and to make specific predictions for future experimental and theoretical research.
* Joint work with James A. Bednar, Yoonsuck Choe, and Joseph Sirosh
Brief Bio.
Risto Miikkulainen is a Professor of Computer Sciences at the University of Texas at Austin. He received an M.S. in Engineering from
the Helsinki University of Technology, Finland, and a Ph.D. in Computer Science from UCLA. His current research includes models of
natural language processing, self-organization of the visual cortex, and evolving neural networks with genetic algorithms. Professor
Miikkulainen is an author of over 180 articles in these research areas, and the books "Computational Maps in the Visual Cortex"
(Springer, 2005), "Lateral Interactions in the Cortex: Structure and Function" (electronic hypertext,
nn.cs.utexas.edu/web-pubs/htmlbook96, 1996), and "Subsymbolic Natural Language Processing" (MIT Press, 1993). He is an editor of the
Machine Learning Journal and Journal of Cognitive Systems Research.
|
| 04/20/05
| Modular Neural Network Models of Functional Brain Imaging
Ransom Winder
|
| 03/07/05
| The Evolution of Arbitrary Computational Processes
Lee Spector, Hampshire College
Abstract
This talk will describe recent advances to the genetic programming paradigm that support the automatic evolution of arbitrary
computational processes. It will focus on techniques involving Push, a programming language designed specifically to support the
evolution of code. Push has a trivial syntax but rich semantics, including facilities for explicit code and control stack manipulation.
These features provide benefits such as automatic code modularization when used in otherwise ordinary genetic programming systems, and
they also support novel "autoconstructive" evolutionary frameworks in which agents construct their own offspring. Several examples will
be presented including evolved programs for standard computer science problems (such as factorial, parity, and sorting), evolved
quantum computing algorithms, and evolved controllers for goal-directed swarms and Quidditch players.
Brief Bio.
Lee Spector is Dean of the School of Cognitive Science and Professor of Computer Science at Hampshire College, and recipient of the NSF
Director's Award for Distinguished Teaching Scholars. His interests include genetic and evolutionary computation, quantum computation,
planning in dynamic environments, artificial intelligence, and neuro- psychology. He recently published the book Automatic Quantum
Computer Programming: A Genetic Programming Approach, Kluwer, 2004.
|
| 02/18/05
| Extending Self-Organizing Particle Systems for Basic Problem Solving
Alejandro Rodriguez
|
| 02/04/05
| A Modular Neural Network of Word and Picture Recognition
Scott Weems
|
|
|
Archive 2004
|
| Date
| Topic & Presenter(s)
| Articles
|
| 12/07/04
| Extending Self-Organazing Particle Systems for Basic Problem Solving
Alejandro Rodriguez
|
| 10/29/04
| Evolving multi-modular networks
JaeYoon Jung
|
| 09/24/04
| Towards intelligent self-assembly
Alex Grushin
| Chris Jones & Maja J. Mataric
From Local to Global Behavior in Intelligent Self-assembly
|
| 05/07/04
| Standard-Scape: An Agent-Based Model of Competition in Technology Markets with Network Externalities
Debbie Heisler
| Judy K. Frels, Debbie Heisler & James A. Reggia
Standard-Scape: An Agent-Based Model of Competition in Technology Markets with Network Externalities.
|
| 04/23/04
| Using Stories to Predict Alternate Futures
Freysi Gudmundsson
| J. Duchan et al.
Deixis in Narrative.
and a recommended web site
|
| 04/02/04
| Generation of Varying Length Temporal Sequences in an Environment from Static Stored Target Trajectories Provided by a Distal
Teacher
Shaun Gittens
| S. Gittens
Dissertation Proposal.
|
| 03/12/04
| Simulated Neuroevolution Produces Networks with Left-Right Asymmetric Processing Speeds and Emergent Lateralization
Alex Grushin
| A. Grushin, J. Reggia
Evolving Processing Speed Asymmetries and Hemispheric Interactions in a Neural Network Model.
|
| 02/20/04
| Adding Memory to Particle Swarm Entities
Ransom Winder
|
| 02/06/04
| A Comparative Study of Hemispheric Interaction Theories
Scott Weems
| S. Weems, J. Reggia
Hemispheric Specialization and Independence for Word Recognition: A Comparison of Three Computational Models.
|
|
|
Archive 2003
|
| 11/07/03
| Collaborating Agents in Artificial Worlds
Alejandro Rodriguez
|
| 10/17/03
| Modeling the Formation of Mirror-Symmetric Maps in Cortex
Reiner Schulz
|
| 10/03/03
| A Neural Network Model of Working Memory
Ransom Winder
| M. A. Tagamets, B. Horwitz [PDF
from publisher]
Integrating electrophysiological and anatomical experimental data to create a large-scale model that simulates a delayed
match-to-sample human brain imaging study.
Cerebral Cortex, vol. 8(4), pp. 310-320, 1998.
|
| 09/12/03
| Descriptive Encoding for Evolving Multi-Modular Neural Networks
Jae-Yoon Jung
|
|
| The Effects of Static Fitness Function Noise Upon the Performance of Genetic Algorithms
Alexander Grushin
| Alexander Grushin
The Effects of Static Fitness Function Noise Upon the Performance of Genetic Algorithms.
Proceedings of the 7th Joint Conference on Information Sciences, pp. 275-278, 2003.
|
| 07/24/03
| Causal Localization of Neural Function: A Fair Attribution of Contribution
Eytan Ruppin, Tel-Aviv University
| S. Kosslyn.
If neuroimaging is the answer, what is the question?
Phil. Trans. R. Soc. Lond. B., vol. 354, pp. 1283-1294, 1999.
R. Aharonov, L. Segev, I. Meilijson, E. Ruppin.
Localization of Function via Lesion Analysis.
Neural Computation., vol. 15(4), pp. 885-914, 2003.
L. Segev, R. Aharonov, I. Meilijson, E. Ruppin.
High-Dimensional Analysis of Autonomous Agents.
Artificial Life., vol. 9(1), pp. 1-20, 2003.
A. Keinan, I. Meilijson, E. Ruppin.
Controlled analysis of neurocontrollers with Informational Lesioning.
Phil. Trans. of the Royal Society of London, A., to appear, Nov 2003.
A. Keinan, C. Hilgetag, I. Meilijson, E. Ruppin.
Fair attribution of contribution: The Shapley value analysis.
Neural Computation.Submitted.
|
| 04/04/03
| Evolving Neural Networks II
Elena Zotenko
| Yao, X. & Liu, Y. (1998)
Making use of population information in evolutionary artificial neural networks.
IEEE Transactions on Systems, Man and Cybernetics (Part B: Cybernetics), vol. 28, pp. 417-425.
|
| 03/07/03
| Coupled Neural Oscillators
James Reggia
| Kimura, H., Akiyama, S. & Sakurama, K. (1999)
Realization of dynamic walking and running of the quadruped using neural oscillators.
Autonomous Robots, vol. 7, pp. 247-258.
Ijspeert, A. J. (2001) A connectionist central pattern generator for the aquatic and terrestrial gaits of a simulated
salamander.
Biological Cybernetics, vol. 84, no. 5, pp. 331-348.
|
| 02/14/03
| Flocking
Alejandro Rodriguez,
David Wang,
Ransom Winder
| Reynolds, C. W. (1987)
Flocks, herds, and schools: a distributed behavioral model.
Computer Graphics (SIGGRAPH '87 Conference Proceedings), vol. 21, no. 4, pp. 25-34.
Reynolds, C. W. (1999)
Steering behaviors for autonomous characters.
Proceedings of the Game Developers Conference 1999 (held in San Jose,California), pp. 763-782, Miller Freeman Game Group, San
Francisco, California.
Ward, C. R., Gobet, F. & Kendall, G. (2001)
Evolving collective behavior in an artificial ecology.
Artificial Life, vol. 7, no. 2, pp. 191-209, MIT Press.
|
|
|
Archive 2002
|
| 11/22/02
| Rule Extraction from Neural Networks
Matt Radio
| Fu, Fu, L. M. (2000)
The Application of certainty factors to neural computing for rule discovery.
IEEE Transactions on Neural Networks, vol. 11, no. 3, pp. 647-657.
|
| 10/25/02
| Evolving Neural Networks
Jae-Yoon Jung
| Poli, R.(1997)
Evolution of graph-like programs with parallel distributed genetic programming.
Genetic Algorithms: Proceedings of the Seventh International Conference , pp. 346-353, Back T. (ed.), Morgan Kaufmann.
|
| 09/27/02
| Neural Modeling of Speech Production
Reiner Schulz
| Guenther, F. H. (1995)
Speech sound acquisition, coarticulation, and rate effects in a neural network model of speech production.
Psychological Review, vol. 102, pp. 594-621.
|
|
|
