Wnek, J. and R.S. Michalski #1994#. Hypothesis driven constructive induction in AQ17HCI: A method and experiments. Machine Learning 14, 139#168.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....is, the prediction accuracy and complexity of constructed decision trees. Essential goal of many scientific researches and experiments is to minimise required computer resources for data classification. One way to develop decision tree induction method is based on construction of new attributes [5]. Constructing new attributes provides a better and faster data classification. The method OC1 [4] constructs new attributes using the linear or polynomial combinations of basic attributes. The OC 1 is known as oblique decision trees method that solves classification exercises with higher accuracy ....

Wnek J. and Michalski R.S. Hypothesis-driven constructive induction in AQ 17-HCi: a method and experiments//in: Machine Learning, 14, 1994 - p. 139 - 168.

....with Jon Oliver and David Hand. I was second author in the first paper and first author in the second paper. The C program to produce the results presented here was jointly written by myself and Jon Oliver. 155 Michalski and Stepp [98] b) COBWEB Fisher [50] c) AQ17 Wnek and Michalski [159], d) AUTOCLASS Cheeseman et al. 34] and (e) Snob Wallace et al. 150, 153] Spath [139, Page 7] defines unsupervised learning (or cluster analysis) in the following way: The objective of cluster analysis is to separate a set of objects into constituent groups (classes, clumps, clusters) ....

J. Wnek and R.S. Michalski. Hypothesis-driven constructive induction in AQ17-HCI: A method and experiments. Machine Learning, 14:139--168, 1994.

....(KDD) it is very often the pre processing stage, namely transformation and reduction of data, that play an important role. A good transformation reduction technique may result in new attributes that are more appropriate for a data mining algorithm. A wide range of constructive induction techniques [6, 7, 9, 15] has been explored aiming at enriching the language of propositional learners [4, 5] The usual way is to add new attributes that are computed from existing ones using a priori given functions. One important class of functions includes linear combinations of numerical attributes. To prevent a ....

Wnek, J. and Michalski, R.S.: Hypothesis-driven Constructive Induction in AQ17HCI: A Method and Experiments. Machine Learning, Vol. 14, No. 2, pp. 139-168, 1994.

.... C) D E F) Figure 2: A univariate tree for (A B C) D E F) There is a group of constructive decision tree learning algorithms, including Fringe (Pagallo and Haussler, 1990) Citre (Matheus and Rendell, 1989) and DCFringe (Yang, Rendell and Blix, 1991) that employ a hypothesis driven strategy (Wnek and Michalski, 1994). These algorithms typically use all or some of conjunction, disjunction and negation as constructive operators, and use decision trees to restrict their new attribute search space. Their control structures are very similar: interleaving a tree learning phase and a new attribute construction ....

....reason, the CI algorithms select new attributes through building a decision tree. All new attributes not used in the decision tree are deleted. FRINGE, DUAL FRINGE, SYMMETRIC FRINGE (Pagallo, 1990) and SFRINGE also use decision trees to select new attributes. Like the CI algorithms, AQ17 HCI (Wnek and Michalski, 1991; 1994) also adopts the hypothesisdriven constructive strategy. AQ17 HCI iterates two processes: rule learning and new attribute construction. In each iteration, rules are generated based on existing attributes, and then new Constructing Conjunctive Attributes using Production Rules Journal of Research ....

[Article contains additional citation context not shown here]

WNEK, J. and MICHALSKI, R.S. (1994): Hypothesis-driven constructive induction in AQ17-HCI: A method and experiments. Machine Learning, 14:139--168.

....reason, the CI algorithms select new attributes through building a decision tree. All new attributes not used in the decision tree are deleted. FRINGE, DUAL FRINGE, SYMMETRIC FRINGE (Pagallo, 1990) and SFRINGE also use decision trees to select new attributes. Like the CI algorithms, AQ17 HCI (Wnek and Michalski, 1991; 1994) also adopts the hypothesisdriven constructive strategy. AQ17 HCI iterates two processes: rule learning and new attribute construction. In each iteration, rules are generated based on existing attributes, and then new Constructing Conjunctive Attributes using Production Rules Journal of ....

WNEK, J. and MICHALSKI, R.S. (1991): Hypothesis-driven constructive induction in AQ17-HCI: A method and experiments. Proceedings of the IJCAI-91 Workshop on Evaluating and Changing Representation in Machine Learning, Sydney, Australia.

....are more representationally powerful than the primitive attributes. Existing constructive induction systems use different strategies for constructing new attributes. There are three primary strategies: the hypothesis driven strategy, the datadriven strategy, and the knowledge driven strategy [Wnek and Michalski, 1994]. Systems with the hypothesis driven strategy analyze hypotheses learned by selective components to discover patterns. Members of this group are algorithms such as Citre 12 Thanks to Thierry Van de Merckt for the suggestion of presenting the problem in this way. 13 [Matheus and Rendell, 1989] ....

....hypotheses learned by selective components to discover patterns. Members of this group are algorithms such as Citre 12 Thanks to Thierry Van de Merckt for the suggestion of presenting the problem in this way. 13 [Matheus and Rendell, 1989] Fringe [Pagallo and Haussler, 1989] and AQ17 hci [Wnek and Michalski, 1994]. Systems with the data driven strategy, such as ID2 of 3 [Murphy and Pazzani, 1991] and AQ17 dci [Bloedorn and Michalski, 1991] find relevant patterns from input data directly, while systems with the knowledge driven strategy apply domain knowledge to create new attributes. This strategy needs ....

[Article contains additional citation context not shown here]

J. Wnek and R.S. Michalski, Hypothesis-driven constructive induction in AQ17-hci: a method and experiments. Machine Learning, 14, 139-168.

....Assuming the original set consists of A 1 ; A 2 ; A n features, these variants can be defined below. Feature construction is a process that discovers missing information about the relationships between features and augments the space of features by inferring or creating additional features [5, 7, 6]. After feature construction, we may have additional m features A n 1 ; A n 2 ; A n m . For example, a new feature A k (n k n m) could be constructed by performing a logical operation on A i and A j from the original set. Another example is: a two dimensional problem (say, A 1 =width and A ....

J. Wnek and R.S. Michalski. Hypothesis-driven constructive induction in aq17-hci: A method and experiments. Machine Learning, 14, 1994.

....and in the representation of examples. When a solution space turns out to be inadequate, representation modification is needed and the modification process typically involves searching for useful new descriptive features (constructive induction) in terms of existing features or attributes. AQ17 [25] of the AQ like family has been developed to implement iterative construction of new attributes based on existing ones. Zheng (1995) has also tried a method called X of N attributes in constructive decision tree construction. Constructive learning has become a strong theme in inductive learning ....

J. Wnek and R.S. Michalski, Hypothesis-Driven Constructive Induction in AQ17-HCI: A Method and Experiments, Machine Learning, 14(1994): 139--168.

....A preprocessing of descriptions would make it possible to determine a hierarchy of generalization of the numerical values. The creation of new values of attributes, as is the case in constructive induction, would make it possible to better account for the similarities between descriptions [18] [26]. Acknowledgements The authors wish to specially thank the anonymous reviewers for their constructive reviews, suggestions and help for writing the final version of this paper. We also would like to thank Lise Fontaine for her careful proofreading of the final version. ....

Wnek J., Michalski R.: Hypothesis-driven constructive induction in AQ17-HCI: a method and experiments. Machine Learning 14(2). (1994). 139-168

....This is equivalent to the binary coding mentioned earlier when the neural net input coding was discussed. 5 Related Work In order to overcome the replication problem and alleviate the fragmentation problem, researchers have suggested various solutions. 16] proposed compound boolean features; [12, 28] employed constructive induction to construct new features defined in terms of existing features; 27] tried global data analysis assessing the value of each partial hypothesis by recurring to all available training data. However, all these solutions are mainly designed for boolean data. The ....

J. Wnek and R.S. Michalski. Hypothesis-driven constructive induction in AQ17HCI: A method and experiments. Machine Learning, 14, 1994.

....case would be when the target concept is actually described by one compound feature. In such a case, we would need only one divide and conquer step to learn the target concept. Feature construction has been proposed as a solution to the problems of repetition, replication and fragmentation [9, 15, 12, 22]. Methods of constructive induction can be classified on the basis of the source of information that is used for searching for the compound features. Many forms of constructive induction exist [10] There is data driven constructive induction (DCI) in which search is based on the analysis of the ....

J. Wnek and R.S. Michalski. Hypothesis-driven constructive induction in aq17-hci: A method and experiments. Machine Learning, 14, 1994.

....new concepts are hard to express in the representation language of decision trees, using only the original attributes. The introduction of the new attributes produces an extension of the representational power and relaxes the language bias of decision tree learning algorithms. Wnek and Michalski [WM94] classify existing Constructive Induction systems into four categories: 34 CHAPTER 1. INTRODUCTION 1. Data Driven constructive induction systems, that analyze and explore the input data, particularly the interrelationships among descriptors used in the examples, and on that basis suggest changes ....

Janusz Wnek and Ryszard S. Michalski. Hypothesis-driven constructive induction in AQ17-HCI: A method and experiments. Machine Learning, 14:139--168, 1994.

....been developed. They perform differently in different domains. Different constructive induction systems use different constructive operators, different theory description languages, different strategies such as a data driven strategy, a hypothesis driven strategy, and a knowledge driven strategy [Wnek and Michalski, 1994], different new attribute evaluation functions, different new attribute search methods, and so on. Constructive operators play an important role in constructive induction. However, all other factors may also affect the performance of a learning system. Each of them might, either positively or ....

....factors may also affect the performance of a learning system. Each of them might, either positively or negatively, contribute to performance differences between different algorithms. Most constructive induction systems such as Fringe [Pagallo, 1990] LFC [Ragavan and Rendell, 1993] and AQ17 hci [Wnek and Michalski, 1994] use conjunction and or disjunction as constructive operators. That is, the constructed attributes are conjunctions or disjunctions of other attributes. A few systems use other constructive operators, for example, M of N [Murphy and Pazzani, 1991; Ting, 1994] mathematical operators such as ....

[Article contains additional citation context not shown here]

J. Wnek and R.S. Michalski, Hypothesis-driven constructive induction in AQ17-hci: a method and experiments. Machine Learning, 14, 139-168.

....almost perfect theories. Therefore constructive induction is only occasionally able to improve the scores marginally. But in every test run the initially induced theory was rewritten into a concise and easily comprehensible form like exemplified by the above given sample rule set. reported in [Wnek Michalski 94] Still this approach does not seem to be able to handle considerable levels of noise. 6.4. EXPERIMENTS 70 Noise First Best 0 63.37 99.51 5 57.36 99.20 10 50.05 50.05 Table 6.4: PARITY 5: accuracies (percentages) for CiPF 2.0 after the first and after the best cycle of induction for various ....

Wnek J., Michalski R.S.: Hypothesis-Driven Constructive Induction in AQ17-HCI: A Method and Experiments, Machine Learning, 14(2), pp.139-168, 1994.

....neither tries to extend ideas from propositional constructive induction to first order logic, nor builds on existing work on feature construction. In this report we try to apply the frameworks for constructive induction by Matheus [Matheus Rendell 89, Matheus 91] and Wnek and Michalski [Wnek Michalski 94] to predicate invention. In the next section, we recall the major reasons for constructive induction in propositional languages. In the third section we review the frameworks for constructive induction by Matheus, Wnek and Michalski. We then give a brief overview of PI methods and the types of ....

....2 Motivations for Constructive Induction In this section we want to recall the motivation for constructive induction in general, since it also applies to predicate invention. Generally, the goal of constructive induction is an increase in accuracy and a decrease in complexity of a hypothesis [Wnek Michalski 94] It is important to include the complexity of the newly defined features or predicates in the complexity of the hypothesis we have to pay a price for defining a large number of new features or predicates. Otherwise we would overfit the potential noise in the data by means of the language, a ....

[Article contains additional citation context not shown here]

Wnek J., Michalski R.S.: Hypothesis-Driven Constructive Induction in AQ17-HCI: A Method and Experiments, in Special Issue on Evaluating and Changing Representation, Machine Learning, 14(2), 1994.

....here for unsupervised learning tasks. 1 INTRODUCTION We discuss the unsupervised learning problem. There are many approaches to unsupervised learning. Within AI there have been systems such as (a) CLUSTER Michalski and Stepp [ 16 ] b) COBWEB Fisher [ 13 ] c) AQ17 Wnek and Michalski [ 29 ] , d) AUTOCLASS Cheeseman et al. 7 ] and (e) Snob Wallace et al. 25; 26 ] which uses MML. Spath [ 23, Page 7 ] defines unsupervised learning (or cluster analysis) in the following way: The objective of cluster analysis is to separate a set of objects into constituent groups (classes, ....

J. Wnek and R.S. Michalski. Hypothesis-driven constructive induction in AQ17-HCI: A method and experiments. Machine Learning, 14:139--168, 1994.

....new attributes is too large to be exhaustively explored. There is a group of constructive decision tree learning algorithms, including Fringe (Pagallo and Haussler, 1990) Citre (Matheus and Rendell, 1989) and DCFringe (Yang, Rendell, and Blix, 1991) that employ a hypothesis driven strategy (Wnek and Michalski, 1994). These algorithms typically use all or some of conjunction, disjunction, and negation as constructive operators, and use decision trees to restrict their new attribute search space. Their control structures are very similar: interleaving a tree learning phase and a new attribute construction ....

....building a decision tree. All new attributes not used in the decision tree are deleted. Fringe, Dual Fringe, Symmetric Fringe (Pagallo, 1990) and SFringe also use decision trees to select new attributes. Some other decision tree learning algorithms use the data driven constructive strategy (Wnek and Michalski, 1994). Instead of using the results of selective induction to guide their new attribute construction, they create new attributes directly from training data when building decision trees. For example, Cart (Breiman et al. 1984) can create one Boolean combination of primitive attributes as a new ....

Wnek, J. and Michalski, R.S. (1994): Hypothesis-driven constructive induction in AQ17-hci: a method and experiments. Machine Learning, 14:139-168.

....that finding reducts is important to simplify decision tables, and hence understanding of the problem; however, when all relative reducts are big, a different decomposition is called for. Example 1 (Multiplexer) Figure 1 depicts a General Logic Diagram (GLD) Michalski, 1978, Thrun et al. 1991, Wnek and Michalski, 1994 ] for a concept. Each instance in the space has exactly one box that is marked with an X if it belongs to the concept and with an O if it does not. Readers familiar with Karnaugh maps may note a resemblance, except that the ordering of attribute values does not conform to the hamming distance ....

Janusz Wnek and Ryszard S. Michalski. Hypothesis-driven constructive induction in AQ17-HCI : A method and experiments. Machine Learning, 14(2):139--168, 1994.

No context found.

Wnek, J. & Michalski, R.S. (1994b). Hypothesis-driven Constructive Induction in AQ17-HCI: A Method and Experiments. Machine Learning, 14, 139-168.

....space for learning. A learning process that consists of two (intertwined) phases, one concerned with the construction of the best representation space, and the second concerned with generating the best hypothesis in the found space is called constructive induction [Mic78] Mic83] [WM94]. An example of a constructive induction program is AQ17 [BWM93] which performs all three types of improvements of the original representation space. In this program, the process of generating new attributes is done by combining initial attributes by mathematical and or logical operators and ....

....the state of a company during selected time instances, etc. Columns in the tables correspond to attributes used to characterize entities associated with the rows. These may be initial attributes, given a priori, or additional ones generated through a process of constructive induction (e.g. [WM94]) Each attribute is assigned a domain and a type. The domain specifies the set of all legal values that the attribute can be assigned in the table. The type defines the ordering (if any) of the values in the domain. For example, the AQ15 learning program [MMHL86] allows four types of attributes: ....

Wnek, J. and Michalski, R.S. Hypothesis-driven Constructive Induction in AQ17-HCI: A Method and Experiments. Machine Learning, 14, pp. 139-168, 1994.

No context found.

Wnek, J. & Michalski, R.S. (1994b). Hypothesis-driven Constructive Induction in AQ17-HCI: A Method and Experiments. Machine Learning, 14, 139-168.

No context found.

Wnek, J. and R.S. Michalski #1994#. Hypothesis driven constructive induction in AQ17HCI: A method and experiments. Machine Learning 14, 139#168.

No context found.

J. Wnek and R.S. Michalski. Hypothesis-driven constructive induction in AQ17-HCI: A method and experiments. Machine Learning, 14:139--168, 1994.

No context found.

J. Wnek and R. S. Michalski. Hypothesis-driven constructive induction in aq17-hci: A method and experiments. Machine Learning, 14:139-168, 1994.

No context found.

J. Wnek and R. Muchalski. Hypothesis-driven constructive induction in aq17-hci - a method and experiments. Machine Learning, 14(2):139--168, 1994.

First 50 documents

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC