• to discuss the need of integrating the symbolic and connectionist approaches;
• to present various approaches to the integration of the symbolic and connection-ist approaches (connectionist symbol processing, hybrid systems, etc.);
• to stimulate students to think about difficult and currently open problems;
• to develop skills in reading scientific papers, writing commentaries on them, giving presentations at a seminar.

• demonstrate familiarity with various ideas for the integration of both approaches;
• debate and discuss theoretical issues and concepts as well as evaluate psychological adequacy of models;
• demonstrate a detailed and critical understanding of research papers – both theoretical and experimental, identify and explain significant developments and contradictions;
• present arguments and evidence coherently in verbal and written form;
• critically evaluate particular approaches to integration.

• guided reading;
• tutor-led and student-led group discussions.

Special Issue of Artificial Intelligence, 46, 1990.

[HLCM] Barnden, J. & Pollack, J. (1991),

High-Level Connectionist Models, Advances in Connectionist and Neural Computation Theory, vol. 1, Norwood, NJ: Ablex Publ. Corp.

[Gap] Dinsmore, J. (ed.),

The Symbolic and Connectionist Paradigms: Closing the Gap, Lawrence Erlbaum, Hillsdale, NJ, 1992.

[HS] Hendler, J. (1989b),

Hybrid Systems (Symbolic/Connectionist), Special Issue of Connection Science, vol. 1(3).

[INSP] Sun, R. & Bookman, L. (1992),

Integrating Neural and Symbolic Processes: The Cognitive Dimension, AAAI Press.

[INSP1] Sun, R. & Bookman, L. (1993),

Special Issue: Integrating Neural and Symbolic Processes. Connection Science, vol. 5 (3&4).

[INSP2] Sun, R. & Bookman, L. (1994),

Computational Architectures Integrating Neural and Symbolic Processes: A Perspective on the State of the Art, Kluwer Academic Publ.

[INSP3] Sun, R. & Alexandre, F. (1995),

Connectionist-Symbolic Integration: From Unified to Hybrid Approaches, IJCAI'95 Workshop Proceedings. (to appear as a book).

Additional readings:
[SESAME] Chown, E. (1992),

An Overview of a Connectionist Cognitive Architecture: The SESAME Approach, In: [INSP]

[SESAME1] Chown, E. & Kaplan (in press),

Active Symbols, Limited Storage and the Power of Natural Intelligence. Behavioral and Brain Sciences.

[Hilario] Hilario, M. (1995),

An Overview of Strategies for Neurosymbolic Integration, In: [INSP3].

[Hilario] Hilario, M., Lallement, Y., Alexandre, F. (1995),

Neurosymbolic Integration: Unified versus Hybrid Approaches. The European Symposium on Artificial NN.

[TRACE] Kaplan, Sonntag, Chown (1991),

Tracing Recurrent Activity in Cognitive Elements (TRACE): A Model of Temporal Dynamics in a Cell Assembly, Connection Science, vol. 3,179-206.

[Lallement1] Lallement, Y. & Alexandre, F. (1995),

Cognitive Aspects of Neurosymboloic Integration, In: [INSP3]

[Lallement2] Lallement, Y., Hilarion, M., Alexandre, F. (1995),

Neurosymbolic Integration: Cognitive Grounds and Computational Strategies. World Conference on the Fundamentals of AI.

[Stark2] Stark, R. (1992),

Restricted Binding in a Strong Hybrid Architecture

[Sun1] Sun, R. (1992),

Connectionist Models of Rule-Based Reasoning, In: [INSP]

[Sun2] Sun, R. (1992),

A Feature-Based Connectionist Inheritance Schema, In: [INSP]

• introduction – discussion of the limitations of both approaches and the need of their integration;
• hybrid systems – overview of various hybrid architectures (architectures com-bining connectionist and symbolic components or mechanisms);
• connectionist symbol processing overview of various attempts to model symbolic processing by connectionist architectures;
• concluding discussions – pros and cons of hybrid models, open problems, re-search trends and directions.

• Fodor, J. & Pylyshyn, Z. (1988) Connectionism and Cognitive Architecture: A Critical Analysis. Cognition, vol. 28
• Harnad, S. (1990) The Symbol Grounding Problem. Physica D, 42, 335-346
• Chalmers, D. (1992) Subsymbolic Computation and the Chinese Room. In: [Gap]

Topic 2: Possible ways of integration: hybrid models (weak integration), connectionist symbol processing (strong integration), implementational and radical connectionism
Required reading:

• Smolensky, P. (1988) On the Proper Treatment of Connectionism. Behavioral and Brain Sciences, 11, 1-74
• Dyer, M. (1991) Symbolic Neuroengineering for Natural Language Processing: A Multilevel Approach. In: [HLCM]
• Schwartz, J. (1992) Who's Afraid of Multiple Realizability?: Functionalism, Reductionism, and Connectionism. In: [Gap]

Additional reading:

• Dinsmore, J. (1992) Thunder in the Gap. In: [Gap]
• Stark, R. (1991) Does Hybrid Mean More than One? Report on the JCI Workshop on Hybrid Models. AISB Quaterly, Special Feature on Hybrid Models, No 78, pp.8-9

Topic 3: Problems in integration: complex structures, temporary associations, sequential processing and control, etc. [Issues], [Problems]
Required reading:

• Barnden, J. (1992) Connectionism, Generalization, and Propositional Attitudes: A Catalogue of Challenging Issues. In: [Gap]
• Barnden, J. & Pollack, J. (1991) Introduction: Problems for High-Level Connectionism. In: [HLCM]

Hybrid Systems

Topic 4: Marker passing and spreading activation: Hendler's approach.
Required reading:

• Hendler, J. (1989) Marker Passing over Microfeatures: Towards a Hybrid Symbolic/Connectionist Model. Cognitive Science, vol. 13, pp. 79-106
• Hendler, J. (1991) Developing Hybrid Symbolic/ Connectionist Models. In: [HLCM]

Additional reading:

• Hendler, J. (1989) On the need for Hybrid Systems. Connection Science, vol. 1(3)

Topic 5: Marker passing and spreading activation: Lange's approach. [ROBIN]
Required reading:

• Lange, T. & Dyer, M. (1989) High-Level Inferencing in a Connectionist Network. Connection Science, vol. 1, pp. 181-217

Topic 6: Shastri's approach.
Required reading:

• Shastri, L. (1991) The Relevance of Connectionism to AI: A Representation and Reasoning Perspective. In: [HLCM]

Topic 7: Micro-Level Hybridization.
Required reading:

• Kokinov, B. (in press) Micro-Level Hybridization in the Cognitive Architecture DUAL. In: Sun, R. & Alexander, F. (eds.) Connectionist-Symbolic Integration. Erlbaum, Hillsdale, NJ
• Kokinov, B. (1994) The Context-Sensitive Cognitive Architecture DUAL. In: Proceedings of the 16th Annual Conference of the Cognitive Science Society. Erlbaum, Hillsdale, NJ
• Kokinov, B. (1994) The DUAL Cognitive Architecture: A Hybrid Multi-Agent Approach. In: A. Cohn (ed.) Proceedings of ECAI'94. John Wiley & Sons, Ltd., London

Topic 8: The Multilevel Dyer's Approach
Required reading:

• Dyer, M. (1990) Distributed Symbol Formation and Processing in Connectionist Networks. JETAI, 2, 215-239
• Dyer, M. (1991) Symbolic Neuroengineering for Natural Language Processing: A Multilevel Approach. In: [HLCM]

Connectionist Symbol Processing

Topic 8: Distributed representation and rules: Touretzky's approach. [DCPS], [BoltzCONS]
Required reading:

• Touretzky, D. & Hinton, G. (1988) A Distributed Connectionist Production System. Cognitive Science, vol. 12
• Touretzky, D. (1990) BoltzCONS: Dynamic Symbol Structures in a Connectionist Network. AI, vol. 46 (1-2)

Topic 9:
Required reading:

• Derthick, M. (1990) Mundane reasoning by settling on a plausible model. AI, vol. 46(1-2)

Topic 10: Part-whole hierarchies: Hinton's approach.
Required reading:

• Hinton, G. (1990) Mapping part-whole hierarchies into connectionist networks. AI, vol. 46 (1-2)

Topic 11: Recursive connectionist representations1: Pollack's approach. [RAAM], [RAAM1]
Required reading:

• Pollack J. (1990) Recursive Distributed Representations. AI, vol. 46 (1-2)
• Blank, D., Meeden, L., Marshall, J. (1992) Exploring the Symbolic/Subsymbolic Continuum: A Case Study of RAAM. In: [Gap]

Topic 12: Recursive connectionist representations2: Elman's Approach
Required reading:

• Elman, J. (1990) Finding Structure in Time. Cognitive Science 14, 179-211
• Elman, J. (1993) Learning and Development in Neural Networks: The Important of Starting Small. Cognition, 48, 71-99.
• Elman, J. (1995) Language Processing. In: Arbib (ed.) Handbook of Brain Theory and NN. MIT Press.
• Elman, J. (1995) Language as a Dynamic System. In: Port, R. & van Gelder, T. (eds.) Mind as Motion. MIT Press.

Topic 13: Radical connectionism: Smolensky's approach.
Required reading:

• Smolensky, P. (1990) Tensor Product Variable Binding and the Representation of Symbolic Structures in Connectionist Systems. AI, vol. 46 (1-2)
• Smolensky, P. & Legendre, G. & Miyata, Y. (1992) Principles for an Integrated Connectionist/Symbolic Theory of Higher Cognition. Technical Report CU-CS-600-92, University of Colorado at Boulder.

Additional reading:

• Smolensky, P. (1994). Computational Theories of Mind. In: A Companion to the Philosophy of Mind. Blackwell Publ.
• Smolensky, P. (1995). Constituent Structure and Explanation in an Integrated Connectionist/Symbolic Cognitive Architecture. In: C. Macdonald & G. Macdonald (Eds.) Connectionism: Debates on Psychological Explanation. vol. 2, Oxford: Basil Blackwell.
• Smolensky, P. (1995). On the Projectable Predicates of Connectionist Psychology: A Case for Belief. In: C. Macdonald & G. Macdonald (Eds.) Connectionism: Debates on Psychological Explanation. vol. 2, Oxford: Basil Blackwell.

Concluding Discussions

Topic 14: Discussion: Representation and Compositional Semantics
Required reading:

• Sharkey, N. (1991) Connectionist Representation Techniques. Artificial Intelligence Review, vol. 5
• Sharkey, N. (1992) The Expanding Role of Superpositional Representations in Hybrid Systems. In: [INSP]
• Feldman, J., Connectionist Representation of Concepts. In: Pfeifer, et al. (eds.) Connectionism in Perspective. Elsevier, 1989
• Touretzky, D. (1991) Connectionism and Compositional Semantics. In: [HLCM] Van Gelder, T. (1990) Compositionality: A Connectionist Variation on a Classical Theme. Cognitive Science, vol. 14

Additional reading:

• Shastri, L. (1991) The Relevance of Connectionism to AI: A Representation and Reasoning Perspective. In: [HLCM]
• Smolensky, P. (1990) Tensor Product Variable Binding and the Representation of Symbolic Structures in Connectionist Systems. AI, vol. 46 (1-2)

Topic 15: Discussion: Hybrid Models: Temporary Bridges or Principle Solutions
Required reading:

• Lange, T. (1992) Hybrid Connectionist Models: Temporary Bridges Over the Gap between the Symbolic and the Subsymbolic. In: [Gap]
• Pollack, J. & Barnden, J. (1991) Conclusion. In: [HLCM] Additional readings: Schwartz, J. (1992) Who's Afraid of Multiple Realizability?: Functionalism, Reductionism, and Connectionism. In: [Gap]
• Dinsmore, J. (1992) Thunder in the Gap. In: [Gap]
• Hendler, J. (1989) On the need for Hybrid Systems. Connection Science, vol. 1(3)

• 20% on each written paper
• 20% on participation in group discussions