Sébastien Hélie,
Ph.D.
|
|
|
Department of Psychology |
|
Phone: (805) 893-7909 Fax : (805) 893-4303 |
|
|
Bldg. 551, Rm. 1504/3227 |
|
- Unsupervised connectionist models
- Computational neuroscience
- Hybrid models
- Creative problem solving (incubation)
- Implicit learning
- Bayesian models
- Mathematical psychology (learning functions, response time distributions)
- Statistics
- Skill acquisition
Collaborators
·
Ron Sun
Professional Activities
ANNOUNCEMENT: CLARION
tutorial
A two-hour tutorial session on the CLARION cognitive architecture
will take place on June 14th 2009 at the International Joint Conference on
Neural Networks in
This tutorial introduces CLARION, a dual-process/dual-representation cognitive architecture that centers on the distinction between explicit and implicit cognitive processes. CLARION is neural-network-based, and composed of four main subsystems: the Action-Centered Subsystem (ACS), the Non-Action-Centered Subsystem (NACS), the Meta-Cognitive Subsystem (MCS), and the Motivational Subsystem (MS). The ACS is mainly for action decision-making. The NACS is a slave system to the ACS and is used to store declarative and episodic knowledge. It is also responsible for reasoning in CLARION. The MS is responsible for determining motivational drive levels. The MCS is responsible for cognitive monitoring and parameter setting in both the ACS and NACS, and makes the goal setting determinations based on drive levels from the MS.
In addition, CLARION is based on two other basic theoretical assumptions: representational differences and learning differences of two types of knowledge: implicit versus explicit. The main difference between these two types of knowledge is accessibility. In each subsystem, the top level contains explicit knowledge (easily accessible) whereas the bottom level contains implicit knowledge (harder to access).
Explicit knowledge is represented using symbolic, localist representations; implicit knowledge is represented using distributed representations. The inaccessible nature of implicit knowledge is captured by distributed representations (in the bottom level), because representational units in a distributed representation are capable of accomplishing tasks but are less individually meaningful. This accords well with the relative inaccessibility of implicit knowledge (as shown by psychology). In contrast, explicit knowledge may be better captured in computational modeling by localist representations (in the top level), in which each unit has a clearer conceptual meaning. This captures the property of explicit knowledge being more accessible.
The second theoretical assumption concerns the different learning processes in the two levels. In the bottom level, implicit associations are learned through gradual trial-and-error learning. In contrast, learning of explicit knowledge is often one-shot and represents the abrupt availability of knowledge . The inclusion and emphasis on bottom-up learning (i.e., the transformation of implicit knowledge into explicit knowledge) is, in part, what distinguishes CLARION from other cognitive architectures. Nevertheless, top-down learning is also carried out in CLARION: Knowledge that is initially explicit can be assimilated into implicit knowledge (to capture proceduralization and automatization found in human psychological data).
CLARION is capable of capturing a wide range of cognitive processes, as well as providing theoretical integration and interpretations of many psychological functions and processes. It has been used to capture numerous tasks. CLARION may also be a useful tool for building cognitively-oriented intelligent systems.
This tutorial presents a detailed description, along with many cognitive simulations, and formal results. Prior exposure to artificial neural networks can be helpful, but prior understanding of cognitive architecture/psychological modeling is not required. This tutorial will enable participants to apply the basic concepts, theories, and computational models of CLARION to their own (cognitively-oriented) work.
For registration, go to: http://www.ijcnn2009.com/
For more information, download the
handout, visit the
CLARION
project website, or contact
Teaching
PSY221A: Design and measurement
Prerequisite: Graduate standing in psychology.
Recommended preparation: A course in calculus.
Experimental design and statistical analysis in psychological research.
Includes basic probability, sampling and distribution theory, hypothesis testing, and estimation. [Website]
COGS-6967/PSYC-4968/MATH-4961: Mathematical foundations of learning in
models of cognition
In the last twenty years, learning has been an important
part of most models of cognition. In many cases, learning is data-driven and
finds its roots in statistical learning. This course covers the most popular
learning procedures used in cognitive science and introduces some important
results and theorems that provide a mathematical foundation for learning in
models of cognition. Understanding the mathematical properties underlying
learning algorithms often reduce the need to use simulations and allows a more
intuitive understanding and thorough explanation of a model’s behavior. More
precisely, this course is divided into four main topics: feature extraction
(i.e., PCA, clustering, entropy, etc.), artificial neural networks (i.e.,
backpropagation, competitive learning, Hebbian learning, landscape analyses,
etc.), Bayesian modeling (i.e., conjugate distributions, parameter learning,
structure learning), and special topics in learning (i.e., theoretical
considerations, formal learning theory, psychological modeling, model
selection). Enrolled students are assumed to have basic understanding of
calculus, mathematical statistics, and linear algebra. [Website]
Texts with an Evaluation Committee
Hélie, S., Proulx, R., & Lefebvre, B. (submitted). Modeling bottom-up learning of explicit knowledge using a Bayesian belief network and a new Hebbian learning rule. Journal of Mathematical Psychology.
Hélie, S. & Sun, R. (submitted). Incubation, insight, and creative problem solving: A unified theory and a connectionist model. Psychological Review.
Hélie, S., Waldschmidt, J.G., & Ashby, F.G. (submitted). Automaticity in rule-based and information-integration categorization. Journal of Experimental Psychology: Learning, Memory, and Cognition.
Hélie, S. & Ashby, G.F. (in press). A neurocomputational model of automaticity and maintenance of abstract rules. Proceedings of the International Joint Conference on Neural Networks. [pdf]
Hélie, S. & Sun, R. (in press). Simulating incubation effects using the Explicit - Implicit Interaction with Bayes factor (EII-BF) model. Proceedings of the International Joint Conference on Neural Networks. [pdf]
Hélie, S. (2008). Energy Minimization in the Nonlinear Dynamic Recurrent Associative Memory. Neural Networks, 21, 1041-1044. [pdf]
Hélie, S. & Sun, R. (2008). Knowledge integration in creative problem solving. In B.C. Love, K. McRae, & V.M. Sloutsky (Eds.) Proceedings of the 30th Annual Meeting of the Cognitive Science Society (pp. 1681-1686). Austin, TX: Cognitive Science Society. [pdf]
Hélie, S., Sun, R., & Xiong, L. (2008). Mixed effects of distractor tasks on incubation. In B.C. Love, K. McRae, & V.M. Sloutsky (Eds.) Proceedings of the 30th Annual Meeting of the Cognitive Science Society (pp. 1251-1256). Austin, TX: Cognitive Science Society. [pdf]
Hélie, S., Wilson, N., & Sun, R. (2008). The CLARION cognitive architecture: A tutorial. In B.C. Love, K. McRae, & V.M. Sloutsky (Eds.) Proceedings of the 30th Annual Meeting of the Cognitive Science Society (pp. 9-10). Austin, TX: Cognitive Science Society. [pdf]
Hélie, S. (2007). Understanding statistical power using noncentral probability distributions: Chi-squared, G-squared, and ANOVA. Tutorials in Quantitative Methods for Psychology, 3, 63-69. [pdf]
Chartier, S.,
Hélie, S., Proulx, R., & Boukadoum, M. (2006). Vigilance procedure
generalization for recurrent associative memories. In R. Sun &
Hélie, S. (2006). An introduction to model selection: Tools and algorithms. Tutorials in Quantitative Methods for Psychology, 2, 1-10. [pdf]
Hélie, S., Chartier, S., & Proulx, R. (2006). Are unsupervised neural networks ignorant? Sizing the effect of environmental distributions on unsupervised learning. Cognitive Systems Research, 7, 357-371. [pdf]
Hélie, S.,
Giguère, G., Cousineau, D., & Proulx, R. (2006). Using knowledge partitioning to investigate the
psychological plausibility of mixtures of experts. Artificial Intelligence Review, 25, 119-138. [pdf]
Hélie, S., Proulx, R., & Lefebvre, B.
(2006). JPEX: A psychologically
plausible Joint
Probability EXtractor. In R.
Sun &
Chartier, S., Hélie, S., Boukadoum,
M., & Proulx, R. (2005). SCRAM: Statistically Converging Recurrent
Associative Memory. Proceedings of the International Joint Conference on
Neural Networks (pp. 723-728). Montréal, QC: IEEE Press. [pdf]
Giguère, G., St-Louis, B., Hélie, S.,
& Harnad, S. (2005). The
role of intra-stimulus variance in perceptual category learning. In B.G. Bara,
L. Barsalou, & M. Bucciarelli (Eds.) Proceedings of the 27th Annual
Meeting of the Cognitive Science Society (pp. 2485). Mahwah, NJ: Lawrence Erlbaum Associates. [pdf]
Hélie, S. & Cousineau, D. (2005). Mixed effects of training on
transfer. In B.G. Bara, L. Barsalou, & M. Bucciarelli (Eds.) Proceedings
of the 27th Annual Meeting of the Cognitive Science Society (pp. 929-934). Mahwah, NJ: Lawrence Erlbaum Associates. [pdf]
Hélie, S., Giguère, G., Cousineau,
D., & Proulx, R. (2005). Are mixtures of experts psychologically plausible? In N. Creany (Ed.) AICS'05:
Proceedings of the 16th Irish Conference on Artificial Intelligence and
Cognitive Science (pp. 61-70). Coleraine, UK: University of Ulster. [pdf]
Giguère, G., Hélie. S., & Cousineau, D.
(2004). Manifeste pour le retour des sciences en psychologie. Revue
Québécoise de Psychologie, 25, 117-130. [pdf]
Hélie, S. (2004). Emergence of Bayesian
structures from recurrent networks. In M. Lovett, C. Schunn, C. Lebiere, & P. Munro (Eds.) Proceedings of the Sixth
International Conference on Cognitive Modelling (pp. 408-409).
Hélie, S., Chartier, S., &
Proulx, R. (2004). Applying fuzzy logic to neural modeling. In M. Lovett, C. Schunn, C. Lebiere, &
P. Munro (Eds.) Proceedings
of the Sixth International Conference on Cognitive Modelling (pp. 352-353).
Cousineau, D., Hélie, S., &
Lefebvre, C. (2003).Testing Curvatures of learning functions on individual
trial and block average data. Behavior Research Methods, Instruments, and
Computers, 35, 493-503. [pdf]
Cousineau, D., Lacroix, G. L., &
Hélie, S. (2003). Redefining the rules: Providing race models with a
connectionist learning rule. Connection Science, 15, 27-43. [pdf]
Hélie, S. & Cousineau, D.
(2003). Testing the equality of learning rates using a linear hypothesis. In R.
Alterman & D. Kirsh (Eds.) Proceedings of the 25th Annual
Meeting of the Cognitive Science Society (pp. 1355).
Book Chapters
Proulx, R. & Hélie, S. (2005). Adaptive
categorization and neural networks. In C. Lefebvre & H. Cohen (Eds.) Handbook
of Categorization in Cognitive Science (pp. 793-815). Oxford: Elsevier.
Other Contributions
Hélie, S. (2007). Modélisation de
l’apprentissage ascendant des connaissances explicites dans une architecture
cognitive hybride. Thèse Doctorale. Montréal,
QC: Université du Québec À Montréal. [pdf]
Hélie, S. (2002). Dissociation de
l’apprentissage des stimuli et de l’apprentissage de la tâche dans le transfert
des habiletés. Mémoire de Maîtrise. Montréal, QC: Université de
Montréal.
Hélie, S. (2002). Les réseaux de
neurones : Est-ce le temps pour une nouvelle génération de modèles? DIRE, 11, 28-29.