UCSC Machine Learning Publications

Worst Case Analysis of On-line Learning by Experts (bibliography)

Worst Case Analysis of On-line Learning by Experts (more complete bibliography in postscript)

UCSC Technical Reports

UCSC Machine Learning FTP site

1999


  1. "Potential Boosters?" N. Duffy and D. Helmbold, Appeared in NIPS'99 as an oral presentation. [Postscript] [BibTex]

  2. "Generalized Connectionist Associative Memory" N. Duffy and A. Jagota, Appeared in IJCAI 99. [Postscript file]

  3. "Using Multiplicative Algorithms to Build Cascade Correlation Networks" N. Duffy. IJCNN99 published by IEEE. [Postscript file] [Abstract]

  4. "Probabilistic Kernel Regression Models" T.S. Jaakkola and D. Haussler. To appear in Proceedings of the 1999 Conference on AI and Statistics. [Compressed Postscript file]

  5. "A Geometric Approach to Leveraging Weak Learners" N. Duffy and D. Helmbold. EuroColt 99 published by Springer Verlag. [Postscript file] [Abstract]

  6. "A Geometric Approach to Leveraging Weak Learners" N. Duffy and D. Helmbold. Submitted to Theory of Computer Science. [Postscript file]

1998


  1. "Exploiting Generative Models in Discriminative Classifiers" T.S. Jaakkola and D. Haussler [Postscript file] [Abstract]

  2. "Tracking the Best Expert," M. Warmuth and M. Herbster, Journal of Machine Learning Vol. 32(2), August 1998. [Postscript file]

  3. "Tracking the Best Disjunction," M. Warmuth and P. Auer, Journal of Machine Learning Vol. 32(2), August 1998. [Postscript file]

  4. "Tracking the Best Regressor," M. Herbster and M. Warmuth. An abstract appeared in Proc. 12th Annu. Conf. on Comput. Learning Theory pp. 24-31, July 1998. [Postscript file]

  5. ``Learning Caller Urgency,'' N. Duffy and D. Helmbold. Appeared in Artificial Intelligence and Cognitive Science 1998. [Postscript file] [Abstract]

  6. ``Tracking a drifting concept in a changing environment.''P.L. Bartlett and D.P. Helmbold. Technical Report UCSC-CRL-98-12, Baskin School of Engineering, University of California, Santa Cruz, 1996, revised 1998.

  7. ``On bayes methods for on-line boolean prediction.''N. Cesa-Bianchi, D.P. Helmbold, and S. Panizza. Algorithmica, 22(1/2):112-137, 1998.

  8. ``Improved lower bounds for learning from noisy examples: an information-theoretic approach.'' C. Gentile and D.P. Helmbold. In Proceedings of the Eleventh Annual Conference on Computational Learning Theory, pages 104-115. ACM Press, July 1998.

  9. ``Comparing gradient-ascent methods for learning mixture distributions.'' D. Herring and D.P. Helmbold. Technical Report UCSC-CRL-98-01, Baskin School of Engineering, University of California, Santa Cruz, 1998.

  10. ``On-line portfolio selection using multiplicative updates.'' D.P. Helmbold, R.E. Schapire, Y. Singer, and M.K. Warmuth. Mathematical Finance, 8(4):325-347, 1998.

1997


  1. "Additive Versus Exponentiated Gradient Updates for Linear Prediction," in Journal of Information and Computation, vol. 132, no. 1, pp. 1-64, January 1997, an extended abstract appeared in STOC 95. M.Warmuth and J. Kivinen [Postscript file]

  2. "The Perceptron Algorithm vs. Winnow: Linear vs. Logarithmic Mistake Bounds When Few Input Variables are Relevant," to appear in the special issue of Artificial Intelligence on Relevance (1997). An extended abstract appeared in COLT 95. M.Warmuth, J. Kivinen and P. Auer [Postscript file]

  3. "Relative Loss Bounds for Multidimensional Regression Problems," an extended abstract was submitted to NIPS 97. M.Warmuth and J. Kivinen [Postscript file]

  4. "Continuous Versus Discrete-Time Non-linear Gradient Descent: Relative Loss Bounds and Convergence," unpublished manuscript. M.Warmuth and A. K. Jagota [Postscript file]

  5. "Worst case prediction over sequences under log loss" M.Opper and D.Haussler (1997) in The Mathematics of Information Coding, Extraction and Distribution, Springer Verlag, Edited by G. Cybenko, D. O'Leary and J. Rissanen. [180k postscript] [Abstract]

  6. "Rigorous Learning Curve Bounds from Statistical Mechanics," Machine Learning , Vol. 25, (1997) pp. 195-236. D.Haussler, Michael Kearns, H. Sebastian Seung and Naftali Tishby [747k postscript] [Abstract]

  7. "Scale sensitive Dimensions, Uniform Convergence, and Learnability," J. ACM 44 (4) 615-631 (1997) N. Alon, S. Ben-David and N. Cesa-Bianchi and D.Haussler [220k postscript] [Abstract]

  8. ``Metric Entropy and Minimax Risk in Classification" D.Haussler and M. Opper, Lecture Notes in Computer Science: Studies in Logic and Computer Science Vol. 1261, 212-235 (1997) Eds. J. Mycielski, G. Rozenberg and A. Salomaa [245k postscript] [Abstract]

  9. "How to Use Expert Advice," J. ACM. Vol.~44 No.~3 (1997) pp. 427-485. Based on University of California, Santa Cruz technical report UCSC-CRL-95-19. N. Cesa-Bianchi, Y. Freund,D.Haussler, D. Helmbold, R. Schapire, and M. Warmuth [218K postscript] [Abstract]

  10. "Mutual Information, Metric Entropy, and Risk in Estimation of Probability Distributions," to appear in Annals of Statistics 25 (6) (Dec. 1997), D.Haussler, Manfred Opper Tech. rep. UCSC-CRL-96-27. [500k postscript] [Abstract]


  11. ``Some label efficient learning results.'' D.P. Helmbold and S. Panizza. In Proceedings of the Tenth Annual Conference on Computational Learning Theory, pages 218-230. ACM Press, July 1997.

  12. ``Predicting nearly as well as the best pruning of a decision tree.'' D.P. Helmbold and R.E. Schapire. Machine Learning, 27(1):51-68, April 1997.

  13. ``A comparison of new and old algorithms for a mixture estimation problem.'' D.P. Helmbold, R.E. Schapire, Y. Singer, and M.K. Warmuth. Machine Learning, 27(1):97-119, April 1997.

1996


  1. ``On-line prediction and conversion strategies.''N. Cesa-Bianchi, Y. Freund, D.P. Helmbold, and M.K. Warmuth. Machine Learning, 25(1):71-110, October 1996.

  2. "Training Algorithms for Hidden Markov Models Using Entropy Based Distance Function," in Advances in Neural Information Processing Systems 9 (NIPS `96), Morgan Kaufmann Publishers, pp. 641-647. M.Warmuth and Y. Singer [Postscript file]

  3. "A General Minimax Result for Relative Entropy," (1996) D.Haussler IEEE Transactions on Information Theory 43(4): 1276-1280 (1997). Also tech. rep. UCSC-CRL-96-26. [154k postscript] [Abstract]

  4. "KDD for Science Data Analysis: Issues and Examples," Proceedings of the Third International Conference on Knowledge Discovery and Data Mining, Portland, Oregon, 1996. D.Haussler, U. Fayyad and P. Stolorz

  5. "Mutual Information, Metric Entropy, and Cumulative Relative Entropy Risk," (1996) D.Haussler and Manfred Opper To appear in Annals of Statistics. [472k postscript] [Abstract]

  6. "Mining Scientific Data," Comm. ACM, 39 (11):1501--1531, November 1996. D.Haussler Usama Fayyad and Paul Stolorz [Abstract]

  7. "Exponentially many local minima for single neurons" Neural Information Processing Systems 1996. Peter Auer, Mark Herbster and Manfred Warmuth [267k postscript] [Abstract]

  8. ``Learning when to trust which experts.'' In Computational Learning Theory: Third European Conference, Eurocolt '97. Springer Verlag, March 1997.

  9. ``A dynamic disk spin-down technique for mobile computing.'' D.P. Helmbold, D.D.E. Long, and B. Sherrod. In Proceedings of the Second Annual ACM International Conference on Mobile Computing and Networking. ACM/IEEE, November 1996.

1995


  1. "Tracking the Best Disjunction," in Proceedings of the 36th Symposium on the Foundations of Comp. Sci., October 95, Milwaukee, pp. 312-321. M.Warmuth and P. Auer [Postscript file of long version]

  2. "General Bounds for Predictive Errors in Supervised Learning," Proceedings of the Workshop on the Theory of Neural Networks: The Statistical Mechanics Perspective, 1995, World Scientific publisher. D.Haussler and Manfred Opper

  3. "Mutual Information and Bayes Methods for Learning a Distribution," Proceedings of the Workshop on the Theory of Neural Networks: The Statistical Mechanics Perspective, Scientific publisher, 1995. D.Haussler and Manfred Opper

  4. "General Bounds on the Mutual Information Between a Parameter and n Conditionally Independent Observations," Proceedings of the Eighth annual Computational Learning Theory Conference (COLT), 1995, Santa Cruz, CA, ACM Press. D.Haussler and Manfred Opper [208k postscript] [Abstract]

  5. "A Generalization of Sauer's Lemma," (1995) Journal Comb. Theory (A), Vol. 71, No. 2, pp. 219-240. D.Haussler and Phil Long

  6. "Bounds for Predictive Errors in the Statistical Mechanics of Supervised Learning," (1995) Physical Review Letters, Vol. 75, No. 20, pp. 3772-3775. M.Opper and D.Haussler [120k postscript] [Abstract]

  7. "Characterizations of Learnability for Classes of {0,...,n}-valued Functions", J. Comp. Sys. Sci. Vol. 50, No. 1, (1995) pp. 74-86. D.Haussler, Shai Ben-David, Nicoló Cesa-Bianchi and Phil Long

  8. "Sphere Packing Numbers for Subsets of the Boolean n-cube with Bounded Vapnik-Chervonenkis Dimension," Journal of Combinatorial Theory (A), Vol. 69, No. 2 (1995) pp. 217-232. D.Haussler

  9. "Tracking the Best Expert" Mark Herbster and Manfred Warmuth, Proceeding of Machine Learning 1995 [524k postscript] [Abstract]

  10. The Perceptron algorithm vs. Winnow: linear vs. logarithmic mistake bounds when few input variables are relevant. Jyrki Kivenen and Manfred K. Warmuth. [417k postscript], [Abstract]

  11. "Using and Combining Predictors that Specialize," in Twentyninth Annual ACM Symposium on Theory of Computing, 1995. M.Warmuth, Y. Freund, R.E. Schapire and Y. Singer [Postscript file]

  12. "Worst-case Loss Bounds for Sigmoided Linear Neurons," in Advances in Neural Information Processing Systems (NIPS `95), Morgan Kaufmann Publishers, pp. 309-315. M.Warmuth, D.P. Helmbold and J. Kivinen [Postscript file]

  13. ``On weak learning.''D.P. Helmbold and M.K. Warmuth. Journal of Computer and System Sciences, 50(3):551-573, June 1995.

1994


  1. Exponentiated Gradient versus Gradient Descent for Linear Predictors. Jyrki Kivenen and Manfred K. Warmuth. [624k postscript]

  2. "Tight Worst-case Loss Bounds for Predicting with Expert Advice," Proceedings of the European Conference on Computational Learning Theory, (EUROCOLT), 1994. (with Jyrki Kivinen and Manfred Warmuth)

  3. "Rigorous Learning Curve Bounds from Statistical Mechanics," Proceedings of the Seventh ACM Conference on Computational Learning Theory, (COLT) , 1994, ACM Press. D.Haussler, Michael Kearns, H. Sebastian Seung, and Naftali Tishby [747k postscript] [Abstract]

  4. "Optimally Parsing a Sequence into Different Classes Based on Multiple Types of Information," Second International Conference on Intelligent Systems in Molecular Biology, Menlo Park, CA. AAAI/MIT Press publisher, August 1994, pp. 369-375. D.Haussler and G.D. Stormo

  5. "Bounds on the Sample Complexity of Bayesian Learning Using Information Theory and the VC Dimension," Machine Learning, 14 (1), (1994) 83-114. D.Haussler, M. Kearns and R. Schapire

  6. "Predicting {0,1}-functions on Randomly Drawn Points," Information and Computation, Vol. 115, No. 2, December 1994, pp. 248-292. D.Haussler, N. Littlestone and M. Warmuth)

  7. "The Weighted Majority Algorithm," in Information and Computation, Vol. 108, No. 2, pp. 212-261 (February 1, 1994). An extended abstract appeared in COLT 89. M.Warmuth and N. Littlestone [Postscript file]

  8. "Sequential Prediction of Individual Sequences Under General Loss Functions," to appear in IEEE Transactions on Information Theory. An extended abstract appeared in EUROCOLT 1994. M.Warmuth, D. Haussler and J. Kivinen [Postscript file]

  9. ``Tracking drifting concepts by minimizing disagreements.'' D.P. Helmbold and P.M. Long. Machine Learning, 14(1):27-45, January 1994.

1993


  1. "Efficient Learning with Virtual Threshold Gates" Wolfgang Maass (TU GRaz) and Manfred K. Warmuth (UC Santa Cruz) [Abstract]

Last modified: Nov 19 1997

maintained by Nigel Duffy