github.com/vlifesystems/rulehunter@v0.0.0-20180501090014-673078aa4a83/examples/csv/monks.txt (about) 1 1. Title: The Monk's Problems 2 3 2. Sources: 4 (a) Donor: Sebastian Thrun 5 School of Computer Science 6 Carnegie Mellon University 7 Pittsburgh, PA 15213, USA 8 9 E-mail: thrun@cs.cmu.edu 10 Web: http://robots.stanford.edu/papers/thrun.MONK.html 11 12 (b) Date: October 1992 13 14 Data Source: 15 UCI Machine Learning Repository [http://archive.ics.uci.edu/ml]. 16 Irvine, CA: University of California, School of Information and 17 Computer Science. 18 http://archive.ics.uci.edu/ml/datasets/MONK%27s+Problems 19 20 21 3. Past Usage: 22 23 - See File: thrun.comparison.ps.Z 24 25 - Wnek, J., "Hypothesis-driven Constructive Induction," PhD dissertation, 26 School of Information Technology and Engineering, Reports of Machine 27 Learning and Inference Laboratory, MLI 93-2, Center for Artificial 28 Intelligence, George Mason University, March 1993. 29 30 - Wnek, J. and Michalski, R.S., "Comparing Symbolic and 31 Subsymbolic Learning: Three Studies," in Machine Learning: A 32 Multistrategy Approach, Vol. 4., R.S. Michalski and G. Tecuci (Eds.), 33 Morgan Kaufmann, San Mateo, CA, 1993. 34 35 4. Relevant Information: 36 37 The MONK's problem were the basis of a first international comparison 38 of learning algorithms. The result of this comparison is summarized in 39 "The MONK's Problems - A Performance Comparison of Different Learning 40 algorithms" by S.B. Thrun, J. Bala, E. Bloedorn, I. Bratko, B. 41 Cestnik, J. Cheng, K. De Jong, S. Dzeroski, S.E. Fahlman, D. Fisher, 42 R. Hamann, K. Kaufman, S. Keller, I. Kononenko, J. Kreuziger, R.S. 43 Michalski, T. Mitchell, P. Pachowicz, Y. Reich H. Vafaie, W. Van de 44 Welde, W. Wenzel, J. Wnek, and J. Zhang has been published as 45 Technical Report CS-CMU-91-197, Carnegie Mellon University in Dec. 46 1991. 47 48 One significant characteristic of this comparison is that it was 49 performed by a collection of researchers, each of whom was an advocate 50 of the technique they tested (often they were the creators of the 51 various methods). In this sense, the results are less biased than in 52 comparisons performed by a single person advocating a specific 53 learning method, and more accurately reflect the generalization 54 behavior of the learning techniques as applied by knowledgeable users. 55 56 There are three MONK's problems. The domains for all MONK's problems 57 are the same (described below). One of the MONK's problems has noise 58 added. For each problem, the domain has been partitioned into a train 59 and test set. 60 61 5. Number of Instances: 432 62 63 6. Number of Attributes: 8 (including class attribute) 64 65 7. Attribute information: 66 1. class: 0, 1 67 2. a1: 1, 2, 3 68 3. a2: 1, 2, 3 69 4. a3: 1, 2 70 5. a4: 1, 2, 3 71 6. a5: 1, 2, 3, 4 72 7. a6: 1, 2 73 8. Id: (A unique symbol for each instance) 74 75 8. Missing Attribute Values: None 76 77 9. Target Concepts associated to the MONK's problem: 78 79 MONK-1: (a1 = a2) or (a5 = 1) 80 81 MONK-2: EXACTLY TWO of {a1 = 1, a2 = 1, a3 = 1, a4 = 1, a5 = 1, a6 = 1} 82 83 MONK-3: (a5 = 3 and a4 = 1) or (a5 /= 4 and a2 /= 3) 84 (5% class noise added to the training set)