Skip to main content
SpringerLink
Log in

Coordinated inductive learning using argumentation-based communication

  • Published:
Autonomous Agents and Multi-Agent Systems Aims and scope Submit manuscript

Abstract

This paper focuses on coordinated inductive learning, concerning how agents with inductive learning capabilities can coordinate their learnt hypotheses with other agents. Coordination in this context means that the hypothesis learnt by one agent is consistent with the data known to the other agents. In order to address this problem, we present A-MAIL, an argumentation approach for agents to argue about hypotheses learnt by induction. A-MAIL integrates, in a single framework, the capabilities of learning from experience, communication, hypothesis revision and argumentation. Therefore, the A-MAIL approach is one step further in achieving autonomous agents with learning capabilities which can use, communicate and reason about the knowledge they learn from examples.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Log in via an institution

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6
Fig. 7
Fig. 8
Fig. 9
Fig. 10

Similar content being viewed by others

Notes

  1. Having a case-base does no imply simply retaining all examples forever; there are techniques for reducing the case-base to a manageable size, as shown in [40, 53].

  2. This is defined in this way mainly for generality and avoiding over-constraining the problem, since the situation where we are interested in finding a single agreed-upon hypothesis for all agents is just a special case of this where we add an additional constraint forcing all the hypotheses to be the same.

  3. Notice that in description logics notation, subsumption is written in the reverse order since it is seen as “set inclusion” of their interpretations. In ML terms, \(A \sqsubseteq B\) means that \(A\) is more general than \(B\), while in description logics it represents the opposite.

  4. Each agent has an individual argumentation model, and there is no “global” argumentation model. Thus, the appraisal on which arguments are deemed as accepted or defeated will vary from one individual agent to another.

  5. A notable exception to that is the work of [20], where they study how a collection of Dung’s style argumentation systems can be merged.

  6. In fact, the model presented in this paper is compatible with Dung’s preferred or grounded extension semantics (which are actually equivalent to dialectical trees when, as in our case, the attack relation has no cycles).

  7. There are other approaches, like categorizers, to mark argument trees; they are discussed later in Sect. 7.

  8. Thus, a rationalist agent \(A_i\) might consider an argument \(\alpha \) to be accepted when marked A, even if it is not \(\tau \)-acceptable for \(A_i\).

  9. In this context, a generalization refinement operator [27] is a function that given a rule condition \(r \in \mathcal {G}\), returns a set of generalizations of the rule condition \(r\), which cover a larger set of examples than \(r\) did. For the experiments reported in this paper, we used the generalization operator defined in [44].

  10. Notice that the previous step updates \(\tau \)-acceptability, assessed based on which examples are covered by each argument, while this step updates the marking of each argument, determined using a marking function over the argumentation tree.

  11. For rationalist agents, the protocol moves to 5 in either case, skipping steps 3 and 4.

  12. Notice that this might have some negative implications in theory, but is useful to ensure termination. In application domains where this causes issues, arguments might be allowed back in the argumentation framework after the agent that withdrew them has received new example arguments; which still ensures termination, since there is a finite number of example-arguments, while avoiding any negative theoretical implications.

  13. Notice that the “Demospongiae” dataset used in this paper is different from the much simpler “sponge” dataset, also from the UCI repository.

  14. The exact datasets used in our experiments can be downloaded from http://sites.google.com/site/santiagoontanonvillar/datasets.

  15. The source code can be found in the following URL: http://sites.google.com/site/santiagoontanonvillar/software.

  16. We didn’t include in the count the rules exchanged in the first step of the protocol, when each agent shares their initial hypotheses with the rest of agents. The reason is that they are not part of the argumentation process: if they already agree in this first step, the cost of argumentation is zero.

  17. As we said before, precision is already good for the individual agents using ABUI  so there is little room from improvement. Precision, however, can vary with \(\tau \) as we will show in Sect. 5.4.

  18. Experiments with the other datasets yielded similar trends. We only report results in the Demospongiae-280 dataset for the sake of space.

  19. In the extreme, when an agent has no cases in the case-base, its recall is 0.00, and its precision is undefined (0 divided by 0).

  20. Higher performance can be achieved by increasing the \(\tau \)-acceptability threshold.

  21. Active learning has the goal to minimize the cost of labelling examples; we do not have this issue here since the solution to examples is already known by the agent that has the example. A scenario closer to active learning would be one in which the agents are not cooperative and giving up information incurs in a payment. Such market-based scenario is beyond the scope of this paper.

  22. Recall that Sect. 5.5 shows the classical teacher/apprentice as a special case on A-MAIL.

References

  1. Aamodt, A., & Plaza, E. (1994). Case-based reasoning: Foundational issues, methodological variations, and system approaches. Artificial Intelligence Communications, 7(1), 39–59.

    Google Scholar 

  2. Aït-Kaci, H., & Podelski, A. (1992). Towards a meaning of LIFE. Tech. Rep. 11, Digital Research Laboratory.

  3. Amgoud, L., & Serrurier, M. (2007). Arguing and explaining classifications. In Proceedings of AAMAS ’07 (pp. 1–7). New York: ACM.

  4. Aras, R., Dutech, A., & Charpillet, F. (2004). Stigmergy in multi agent reinforcement learning. In Proceedings of 4th hybrid intelligent systems (pp. 468–469). Los Alamitos: IEEE Computer Society.

  5. Armengol, E., & Plaza, E. (2000). Bottom-up induction of feature terms. Machine Learning Journal, 41(1), 259–294.

    Article  MATH  Google Scholar 

  6. Bache, K., & Lichman, M. (2013). UCI machine learning repository. http://archive.ics.uci.edu/ml.

  7. Besnard, P., & Hunter, A. (2001). A logic-based theory of deductive arguments. Artificial Intelligence, 128(1), 203–235.

    Article  MATH  MathSciNet  Google Scholar 

  8. Bourgne, G., El Fallah Segrouchni, A., & Soldano, H. (2007). SMILE: Sound multi-agent incremental learning. In Proceedings of AAMAS ’07 (pp. 239:1–239:8). New York: ACM.

  9. Bourgne, G., Soldano, H., & Fallah-Seghrouchni, A. E. (2010). Learning better together. In Proceedings of ECAI’10. Frontiers in artificial Intelligence and applications (Vol. 215, pp. 85–90). Amsterdam: IOS Press.

  10. Bowling, M., & Veloso, M. M. (2003). Simultaneous adversarial multi-robot learning. In Proceedings of IJCAI-03 (pp. 699–704). Edmonton: Morgan Kaufmann.

  11. Bowling, M. H., & Veloso, M. M. (2002). Multiagent learning using a variable learning rate. Artificial Intelligence, 136(2), 215–250.

    Article  MATH  MathSciNet  Google Scholar 

  12. Brazdil, P. B., & Torgo, L. (1990). Knowledge acquisition via knowledge integration. In B. Wielinga, J. Boose, B. Gaines, G. Schreiber, & M. van Someren (Eds.), Current trends in knowledge acquisition. Amsterdam: IOS Press.

  13. Carlson, A., Betteridge, J., Kisiel, B., Settles, B., Hruschka, Jr., E. R., & Mitchell, T. M. (2010). Toward an architecture for never-ending language learning. In AAAI.

  14. Carpenter, B. (1991). Typed feature structures: An extension of first-order terms. In V. Saraswat & K. Ueda (Eds.), Logic programming: Proceedings of the 1991 international symposium (pp. 187–201). Cambridge: The MIT Press.

  15. Carpenter, B. (1992). The logic of typed feature structures. Cambridge tracts in theoretical computer science. Cambridge: Cambridge University Press.

  16. Caruana, R. (1997). Multitask learning. Machine Learning, 28(1), 41–75.

    Article  MathSciNet  Google Scholar 

  17. Chesñevar, C. I., Simari, G. R., & Godo, L. (2005). Computing dialectical trees efficiently in possibilistic defeasible logic programming. In Proceedings of LPNMR’05. Lecture notes in computer science (Vol. 3662, pp. 158–171). Heidelberg: Springer.

  18. Clark, P., & Niblett, T. (1989). The CN2 induction algorithm. Machine Learning, 3, 261–283.

    Google Scholar 

  19. Cohn, D. A., Ghahramani, Z., & Jordan, M. I. (1994). Active learning with statistical models. In G. Tesauro, D. Touretzky, & T. Leen (Eds.), Proceedings of NIPS’94 (pp. 705–712). Cambridge: The MIT Press.

  20. Coste-Marquis, S., Devred, C., Konieczny, S., Lagasquie-Schiex, M. C., & Marquis, P. (2007). On the merging of Dung’s argumentation systems. Artificial Intelligence, 171, 730–753.

    Article  MATH  MathSciNet  Google Scholar 

  21. Davies, W., & Edwards, P. (1995). Distributed learning: An agent-based approach to data-mining. In ICML ’95 workshop on agents that learn from other agents.

  22. Dung, P. M. (1995). On the acceptability of arguments and its fundamental role in nonmonotonic reasoning, logic programming and n-person games. Artificial Intelligence, 77(2), 321–357.

    Article  MATH  MathSciNet  Google Scholar 

  23. Dunne, P. E., Hunter, A., McBurney, P., Parsons, S., & Wooldridge, M. (2009). Inconsistency tolerance in weighted argument systems. In C. Sierra, C. Castelfranchi, K. S. Decker, & J. S. Sichman (Eds.), Proceedings of AAMAS ’09, IFAAMAS (pp. 851–858), Taipei.

  24. Hirsh, H. (1989). Incremental version-space merging: A general framework for concept learning. Ph.D. Thesis, Stanford University, Stanford, CA.

  25. Hu, J., & Wellman, M. P. (1998) Multiagent reinforcement learning: Theoretical framework and an algorithm. In Proceedings of ICML ’98 (pp. 242–250). San Francisco: Morgan Kaufmann.

  26. Karunatillake, N. C., Jennings, N. R., Rahwan, I., & McBurney, P. (2009). Dialogue games that agents play within a society. Artificial intelligence, 173(9), 935–981.

    Article  Google Scholar 

  27. van der Laag, P. R. J., & Nienhuys-Cheng, S. H. (1994). Existence and nonexistence of complete refinement operators. In Proceedings of ECML-94. Lecture notes in computer science (Vol. 784, pp. 307–322). Berlin: Springer.

  28. Larson, J., & Michalski, R. S. (1977). Inductive inference of VL decision rules. SIGART Bulletin, 63, 38–44.

    Article  Google Scholar 

  29. Lavrač, N., & Džeroski, S. (1994). Inductive logic programming. Techniques and applications. New York: Ellis Horwood.

  30. Leake, D. B., & Ram, A. (Eds.). (1995). Goal-driven learning. Cambridge: The MIT Press.

  31. Leake, D. B., & Sooriamurthi, R. (2001). When two case bases are better than one: Exploiting multiple case bases. In Proceedings of ICCBR’01. Lecture notes in computer science (Vol. 2080, pp. 321–335). Berlin: Springer.

  32. Leake, D. B., & Sooriamurthi, R. (2002). Managing multiple case bases: Dimensions and issues. In Proceeding of FLAIRS’02 (pp. 106–110). Menlo Park: AAAI Press.

  33. Littman, M. L. (1994). Markov games as a framework for multi-agent reinforcement learning. In Proceedings of ICML-94 (pp. 157–163). San Francisco: Morgan Kaufmann.

  34. Manning, C., Raghavan, P., & Schutze, M. (2009). Probabilistic information retrieval. Cambridge: Cambridge University Press.

    Google Scholar 

  35. McGinty, L., & Smyth, B. (2001) Collaborative case-based reasoning: Applications in personalized route planning. In Proceedings of ICCBR’01. Lecture notes in computer science (Vol. 2080, pp. 362–376). Berlin: Springer.

  36. Michie, D., Muggleton, S., Page, D., & Srinivasan, A. (1994). To the international computing community: A new East-West challenge. Tech. rep., Oxford University Computing Laboratory, Oxford. ftp://ftp.comlab.ox.ac.uk/pub/Packages/ILP/trains.tar.Z.

  37. Modi, P. J., & Shen, W. M. (2001). Collaborative multiagent learning for classification tasks. In J. P. Müller, E. Andre, S. Sen, & C. Frasson (Eds.), Proceedings of ICAA’01 (pp. 37–38). New York: ACM Press.

  38. Mozina, M., Zabkar, J., & Bratko, I. (2007). Argument based machine learning. Artificial Intelligence, 171(10–15), 922–937.

    Article  MATH  MathSciNet  Google Scholar 

  39. Ontañón, S., Dellunde, P., Godo, L., & Plaza, E. (2012). A defeasible reasoning model of inductive concept learning from examples and communication. Artificial intelligence, 193, 129–148.

    Article  MATH  MathSciNet  Google Scholar 

  40. Ontañón, S., Plaza, E. (2004). Justification-based selection of training examples for case base reduction. In J. F. Boulicaut, F. Esposito, F. Giannotti, & D. Pedreschi (Eds.), Machine learning: ECML 2004. Lecture notes in artificial intelligence (Vol. 3201, pp. 310–321). Berlin: Springer.

  41. Ontañón, S., & Plaza, E. (2007). Learning and joint deliberation through argumentation in multiagent systems. In E. H. Durfee, M. Yokoo, M. N. Huhns, & O. Shehory (Eds.), Proceedings of AAMAS’07 (pp. 971–978). Honolulu: IFAAMAS.

  42. Ontañón, S., & Plaza, E. (2010) Concept convergence in empirical domains. In B. Pfahringer, G. Holmes, & A. G. Hoffmann (Eds.), Discovery science. Lecture notes in computer science (Vol. 6332, pp. 281–295). Berlin: Springer.

  43. Ontañón, S., & Plaza, E. (2010). Towards argumentation-based multiagent induction. In Proceedings of the 2010 conference on ECAI 2010: 19th European conference on artificial intelligence (pp. 1111–1112). Amsterdam: IOS Press.

  44. Ontañón, S., & Plaza, E. (2012). Similarity measures over refinement graphs. Machine Learning, 87(1), 57–92.

    Article  MATH  MathSciNet  Google Scholar 

  45. Plaza, E., Arcos, J. L., & Martín, F. (1997) Cooperative case-based reasoning. In G. Weiss (Ed.), Distributed artificial intelligence meets machine learning. Learning in multi-agent environments. Lecture notes in artificial intelligence (Vol. 1221, pp. 180–201). Berlin: Springer.

  46. Plaza, E., & Ontañón, S. (2006). Learning collaboration strategies for committees of learning agents. Autonomous Agents and Multi-Agent Systems, 13(3), 429–461.

    Article  Google Scholar 

  47. Prakken, H. (2005). Coherence and flexibility in dialogue games for argumentation. Journal of Logic and Computation, 15, 1009–1040.

    Article  MATH  MathSciNet  Google Scholar 

  48. Prassad, M. V. N., Lesser, V. R., & Lander, S. (1995). Retrieval and reasoning in distributed case bases. Tech. rep., UMass Computer Science Department.

  49. Provost, F. J., & Hennessy, D. (1996). Scaling up: Distributed machine learning with cooperation. In W. J. Clancey & D. S. Weld (Eds.), Proceedings of AAAI’96 (pp. 74–79). Menlo Park/Cambridge: AAAI Press/The MIT Press.

  50. Quinlan, J. R. (1990). Learning logical definitions from relations. Machine Learning, 5, 239–266.

    Google Scholar 

  51. Rotstein, N. D., Moguillansky, M. O., & Simari, G. R. (2009). Dialectical abstract argumentation: A characterization of the marking criterion. In C. Boutilier (Ed.), Proceedings of IJCAI’09 (pp. 898–903). Menlo Park: AAAI Press.

  52. Sian, S. S. (1991). Extending learning to multiple agents: Issues and a model for multi-agent machine learning (MA-ML). In Y. Kodratoff (Ed.), Machine learning—EWSL-91. Lecture notes in computer science (Vol. 482, pp. 440–456). Berlin: Springer.

  53. Smyth, B., & Keane, M. T. (1995). Remenbering to forget: A competence-preserving case delection policy for case-based reasoning systems. In Proceedings of IJCAI-95 (pp. 377–382).

  54. Stone, P., & Sen, S. (Eds.). (2000). In Proceedings of AGENTS-2000/ECML-2000 joint workshop on learning agents, 3 June 2000, Barcelona.

  55. Stone, P., & Veloso, M. M. (1998). Towards collaborative and adversarial learning: A case study in robotic soccer. International Journal of Human-Computer Studies, 48(1), 83–104.

    Article  Google Scholar 

  56. Stone, P., & Veloso, M. M. (2000). Multiagent systems: A survey from a machine learning perspective. Autonomous Robots, 8(3), 345–383.

    Article  Google Scholar 

  57. Thimm, M., & Kern-Isberner, G. (2008). A distributed argumentation framework using defeasible logic programming. In Computational models of argument: Proceedings of COMMA 2008. Frontiers in artificial intelligence and applications (Vol. 172, pp. 381–392). Amsterdam: IOS Press.

  58. Wardeh, M., Bench-Capon, T. J. M., & Coenen, F. (2009). PADUA: A protocol for argumentation dialogue using association rules. Artificial Intelligence in Law, 17(3), 183–215.

    Article  Google Scholar 

Download references

Acknowledgments

Research partially funded by the projects Next-CBR (TIN2009-13692-C03-01) and Cognitio (TIN2012-38450- C03-03) [both co-funded with FEDER], Agreement Technologies (CONSOLIDER CSD2007-0022), and by the Grants 2009-SGR-1433 and 2009-SGR-1434 of the Generalitat de Catalunya.

Author information

Authors and Affiliations

  1. Computer Science Department, Drexel University, Philadelphia, PA , 19104, USA

    Santiago Ontañón

  2. IIIA (Artificial Intelligence Research Institute), CSIC (Spanish Council for Scientific Research), Campus UAB, 08193 , Bellaterra, Catalonia, Spain

    Enric Plaza

Authors
  1. Santiago Ontañón
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Enric Plaza
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Santiago Ontañón.

Rights and permissions

Reprints and permissions

About this article

Cite this article

Ontañón, S., Plaza, E. Coordinated inductive learning using argumentation-based communication. Auton Agent Multi-Agent Syst 29, 266–304 (2015). https://doi.org/10.1007/s10458-014-9256-2

Download citation

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s10458-014-9256-2

Keywords

Search

Navigation