Abstract
In order for engineered systems to produce behaviors that achieve esthetic goals, one requires objective functions that accurately represent potentially subjective, human preferences as opposed to a priori given objectives. Starting from a collection of empirical, pairwise comparisons, we approach this issue by developing objective functions that are compatible with the expressed preferences. In addition, robust estimators for global optimizers to these functions are derived together with graph-theoretic simplification methods for the resulting systems of constraints and a limited memory asymptotic observer that finds a globally optimal alternative (e.g., motion). Two examples are presented involving the comparison of apples and oranges, and of human and synthetic motions.
Keywords
This is a preview of subscription content, log in via an institution.
Buying options
Tax calculation will be finalised at checkout
Purchases are for personal use only
Learn about institutional subscriptionsNotes
- 1.
To the extent that a distinction is made between “instances” and “alternatives,” it is that “instances” are the points that were shown to human judges, whereas “alternatives” may also include other points in the space besides those that were seen.
References
Abbeel P, Ng AY (2004) Apprenticeship learning via inverse reinforcement learning. In: Proceedings of the international conference on machine learning
Agarwal S (2006) Ranking on graph data. ICML, ACM Press, pp 25–36
Aho AV, Garey MR, Ullman JD (1972) The transitive reduction of a directed graph. SIAM J Comput 1:131–137
Aiolli F, Sperduti A (2004) Learning preferences for multiclass problems. Advances in neural information processing systems 17, MIT Press, pp 17–24
Aizerman M, Braverman E, Rozonoer L (1964) Theoretical foundations of the potential function method in pattern recognition learning. Autom Remote Control 25:821–837
Bahamonde A, Diez J, Quevedo J, Luances O, del Coz J (2007) How to learn consumer preferences from the analysis of sensory data by means of support vector machines. Trends Food Sci Technol 18:20–28
Baird B (1965) The art of the puppet. Mcmillan Company, New York
Black J, Iyasere E, Wagner J (2011) Creation of a driver preference objective metric to evaluate ground vehicle steering systems. In: American control conference
Boser B, Guyon I, Vapnik V (1992) An training algorithm for optimal margin classiers. In: Proceedings of the fifth annual workshop on computational learning theory, pp 144–152
Callahan P (1994) Puppets, stop the flap!: the Arpeggio method of mouth puppet manipulation. Arpeggio, Atlanta
Chu W, Ghahramani Z (2005) Preference learning with gaussian processes. In: Proceedings of the 22nd international conference on machine learning, New York, pp 137–144
Chu W, Keerthi SS (2007) Support vector ordinal regression. Neural Comput 15(9):2227–2254
Cohen WW, Schapire RE, Singer Y (1999) Learning to order things. J Artif Intell Res 10:243–270
Coppersmith D, Winograd S (1987) Matrix multiplication via arithmetic progressions. In STOC ’87: Proceedings of the nineteenth annual ACM symposium on theory of computing, New York, pp 1–6
Cortes C, Mohri M, Rastogi A (2007) Magnitude-preserving ranking algorithms. In: Proceedings of the twenty-fourth international conference on machine learning
Canudas de Wit C (2005) Fun-to-drive by feedback. Eur J Control: Fundam issues Control 11/4–5:353–383
Diez J, Jose del Coz J, Luaces O, Bahamonde A (2008) Clustering people according to their preference criteria. Expert Syst Appl 34:1274–1284
Egerstedt M, Murphey T, Ludwig J (2007) Motion Programs for Puppet Choreography and Control. Springer-Verlag, Berlin, pp 190–202
Engler L, Fijan C (1973) Making puppets come alive. Taplinger Publishing Company, New York
Fiechter CN, Rogers S (2000) Learning subjective functions with large margins. In: Proceedings of the seventeenth international conference on machine learning, pp 287–294
Fischer MJ, Meyer AR (1971) Boolean matrix multiplication and transitive closure. In: Twelfth annual symposium on switching and automata theory, pp 129–131
Herbrich R, Graepel T, Bollmann-Sdorra P, Obermayer K (1998) Supervised learning of preference relations. In: Proceedings FGML-98, German national workshop on machine learning, pp 43–47
Herbrich R, Graepel T, Obermayer K (2000) Large margin rank boundaries for ordinal regression. Advances in large margin classifiers. MIT Press, Cambridge
Hullermeier E, Frankranz J, Cheng W, Brinker K (2008) Label ranking by learning pair-wise preferences. Artif Intell 172:1897–1917
Jiang X, Lim L-H, Yao Y, Ye Y (2011) Statistical ranking and combinatorial hodge theory. Math Program 127:203–244
Joachims T (2002) Optimizing search engines using click-through data. In: ACM SIGKDD conference on knowledge discovery and data mining (KDD), pp 133–142
Johnson E, Murphey T (2007) Dynamic modeling and motion planning for marionettes: rigid bodies articulated by massless strings. In: IEEE international conference on robotics and automation, pp 330–335
Kingston P, Egerstedt M (2009) Comparing apples and oranges through partial orders: an empirical approach, In: American control conference
Kingston P, Egerstedt M (2011) Motion preference learning. In: American control conference
Munro I (1971) Efficient determination of the strongly connected components and the transitive closure of a graph
Platt JC (1998) Sequential minimal optimization: a fast algorithm for training support vector machines. Advances in kernel methods. MIT Press, Cambridge
Qiao Q, Beling PA (2011) Inverse reinforcement learning with gaussian process. In: American control conference
Roser A (1979) Gustaf Und Sein Ensemble: Beschreibungen eines Puppenspielers. Bleicher Verlag, Gerlingen
Strassen V (1969) Gaussian elimination is not optimal. Numer Math 13:354–356
Syed U, Bowling M, Schapire RE (2008) Apprenticeship learning using linear programming. In: Proceedings of the international conference on machine learning
Acknowledgments
This work was supported by the U.S. National Science Foundation through Creative IT Grant #0757317. The human-study was conducted within the Georgia Institute of Technology, Institute Review Board Protocol H08162 - “Perceived Similarity Study.” We would like to thank Akhil Bahl for his assistance in producing the motion capture data used in the synthetic amoeba experiment.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2014 Springer International Publishing Switzerland
About this chapter
Cite this chapter
Kingston, P., von Hinezmeyer, J., Egerstedt, M. (2014). Metric Preference Learning with Applications to Motion Imitation. In: LaViers, A., Egerstedt, M. (eds) Controls and Art. Springer, Cham. https://doi.org/10.1007/978-3-319-03904-6_1
Download citation
DOI: https://doi.org/10.1007/978-3-319-03904-6_1
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-319-03903-9
Online ISBN: 978-3-319-03904-6
eBook Packages: EngineeringEngineering (R0)