Utgoff, P.: Shift of Bias for Inductive Concept Learning, in: Machine Learning: an artificial intelligence approach (R. Michalski, J. Carbonell, T. Mitchell, Eds.), vol. II, Morgan Kaufmann, Los Altos, CA, 1986, 107--148.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....a concept covering C. That is to say, it is a function which returns a level n property if the input is in C, and undefined otherwise. The function effectively partitions the set of all level n 1 entities into two subsets and this is precisely, the behaviour of a computational concept (cf. [3,4]) An example may help here. Consider the black card entity again. In terms of our framework, this is a (singleton) collection of playing cards which exhibit certain properties; i.e. it is a card which has certain properties (namely, spadeness or club ness) In a hierarchical representation the ....

Utgoff, P. (1986). Shift of bias for inductive concept learning. In R. Michalski, J. Carbonell and T. Mitchell (Eds.), Machine Learning: An Artificial Intelligence Approach: Vol II (pp. 107-148). Los Altos: Morgan Kaufmann.

....all unseen cases (weighted according to T) We provide a precise definition of generalization accuracy later in Equation 5. The generalization performance of 2 is 79( S) 6A( S) 0.5. Throughout this paper, the term learner (or learning algorithm) is equivalent to the bias [Mitchell, 1980; Utgoff, 1986] used by to gener alize from 0 to unseen cases, and V should be read as for all learning biases. The cIGP states that in a classification problem, for any learning algorithm the total generalization performance over all learning situations is zero. v [6v( s) v [6x( s) 0.5] 0 v ....

Utgoff, P. E. 1986. Shift of bias of induc- tive concept learning. In Michalski, R.S.; Carbonell, J.G.; and Mitchell, T.M., editors 1986, Machine Learning: An Artificial Intelligence Approach, Vol II. Morgan Kaufmann Publishers. 107 148.

....wrong results. Chosing an appropriate set of auxiliary features to use when performing classification is a difficult problem, usually described as constructive induction. The approaches developed so far can be viewed as belonging to one of two classes. Analytical approaches [Drastal et A1. 1989, Utgoff 1986] generate new features by using a domain specific theory. The usefulness of the features derived depends on the amount of appropriate domain knowledge available. If there exists good domain knowledge this may be an effective method to generate an adequate set of auxiliary features. Empirical ....

P. E. Utgoff "Shift of Bias for Inductive Concept Learning" in Machine Learning: An Artificial Intelligence Approach, 1986, Morgan Kaufmann.

....l m l (p q) n Gamma1 . An analogous logarithmic lower bound can be obtained using results derived by Ehrenfeucht and colleagues [18, 26] A similar analysis can be found in [7] The key idea underlying this analysis is to make explicit two levels of learning: a meta level and a base level [48, 53, 71]. The base level learning problem is the problem of learning functions, just like regular supervised learning. The meta level learning problem is the problem of learning properties of functions, i.e. learning entire function spaces. Learning at the meta level bears close resemblance to baselevel ....

....(declarative bias) or straight program code (procedural bias) the same representation may be employed for both bias and learned functions. Consequently, knowledge acquired in one task may be used as bias in another. Examples of learning systems that modify declarative bias are SOAR [29] STABB [70, 71], inductive logic programming [17, 41, 47] theory revision [37] and RALPH [51] Procedural bias is learned in genetic programming [15, 27, 28, 64, 65] and in an approach by Schmidhuber and colleagues [52, 53, 54] Notice that rules and program code can be viewed as partially defined functions ....

P. E. Utgoff. Shift of bias for inductive concept learning. In R.S. Michalski, J.G. Carbonell, and T.M. Mitchell, editors, Machine Learning: An Artificial Intelligence Approach, Volume II. Morgan Kaufmann, 1986.

....inductive learning means more than parameter estimation; it means constructing the model itself. This difficult goal has taken several forms in different research fields, including change of representation, variable bias, adaptable network structure, and non parametric statistics (Schlimmer, 1987; Utgoff, 1986; Rendell, Seshu, Tcheng, 1987; Barron Barron, 1988; Devroye Gyorfi, 1985; Geman, Bienenstock, Doursat, 1992) Induction involves the discovery of regularities in data and knowledge structures. The complexity of the existing regularities determines the complexity of the inductive process. ....

Utgoff, P. E. (1986). Shift of bias for inductive concept learning. In Michalski, R., Carbonell, J., & Mitchell, T. (Eds.), Machine Learning: An Artificial Intelligence Approach, Vol. II, chap. 5, pp. 107--148. Morgan Kaufmann Publishers, Inc., Los Altos, CA.

....discourse. Each given representational notation manifests a particular representational guidance, expressing certain aspects of one s knowledge better than others do. The concept of representational guidance is borrowed from artificial intelligence, where it is called representational bias [21]. The phrase guidance is adopted here to avoid the negative connotation of bias. The phrase knowledge unit will be used to refer generically to components of knowledge one might wish to represent, such as hypotheses, statements of fact, concepts, relationships, rules, etc. Representational ....

....based on differences between representational notations. 2. 1 Representational notations bias learners towards particular ontologies The first hypothesis claims that important guidance for learning interactions comes from ways in which a representational notation limits what can be represented [15, 21]. A representational notation provides a set of Figure 1 Representational Guidance To appear in proceedings of International Conference on Computers in Education, November21 24, 2000, Taipei, Taiwan 3 primitive elements out of which representational artifacts are constructed. These primitive ....

Utgoff, P. (1986). Shift of bias for inductive concept learning. In R. Michalski, J. Carbonell, T. Mitchell (Eds.) Machine Learning: An Artitificial Intelligence Approach, Volume II (pp. 107-148). Los Altos: Morgan Kaufmann.

....is 2 100 , or over 10 30 , equal to the number of possible functions, which is equal to the number of subsets that can be formed from the 100 possible instances. Clearly, to successfully search for any nontrivial function requires some method of limiting the number of hypotheses. Biases (Utgoff, 1986) focus a learning method s search, forcing it to consider only hypotheses of a particular form. For instance, a neural network is itself a structural bias, upon which algorithms such as backpropagation operate. Decision rules are also structural biases. Biases transform the problem of learning a ....

Utgoff, P. E. 1986. Shift of bias for inductive concept learning. In Machine learning: An artificial intelligence approach, vol. II, eds. R. S. Michalski, J. G. Carbonell, and T. M. Mitchell, 107148. Los Altos, Calif.: Morgan Kaufmann.

....correctness, effectiveness, noise sensitivity, robustness, etc. These quality criteria allow one to distinguish between different methods for, for example, concept learning that learn qualitatively different descriptions and concept types 16 . They capture at an abstract level, a type of bias [Utgoff, 1984] implicit in a learning method i.e. its preference for one kind of generalization or another. The appropriate bias must be related to characteristics of the environment in which the problem solving activity, to which a learning goal relates, occurs: its criticality, the diversity of problems that ....

Utgoff, P. (1984). Shift of Bias for Inductive Concept Learning. PhD thesis, Rutgers University, Rutgers University.

....to construct a small set of new attributes from primitive attributes such that theories built using the new attributes are more accurate and concise than those created directly using the primitive attributes. Note that people also discuss this issue from different points of view. For example, Utgoff [1986] views the construction of new attributes as an automatic shift of language bias. The principal differences of these constructive induction systems in terms of generating new attributes are that different constructive operators such as conjunction, disjunction, M of N, and multiplication are ....

P.E. Utgoff, Shift of bias of inductive concept learning. In R.S. Michalski, J.G. Carbonell, and T.M. Mitchell (eds.), Machine Learning: An Artificial Intelligence Approach (Vol. II), San Mateo, CA: Morgan Kaufmann, 107-148.

.... continual learning , learning from invariances , learning by knowledge transfer etc. would not be possible, and experience with previous problems could not help to sensibly adjust the prior distribution of solution candidates in the search space for a new problem (shift of inductive bias, e.g. [45]) To repeat: an interesting feature of incremental self improvement is that its theoretical potential for exploiting environmental regularities, if there are any, exceeds the corresponding potential of previous learning systems. 3 Illustrative Experiments The main purpose of this paper was to ....

P. Utgoff. Shift of bias for inductive concept learning. In Machine Learning, volume 2. Morgan Kaufmann, Los Altos, CA, 1986.

....in Heskes (1998) 151 BAXTER Early machine learning work. In Rendell, Seshu, and Tcheng (1987) VBMS or Variable Bias Management System was introduced as a mechanism for selecting amongst different learning algorithms when tackling a new learning problem. STABB or Shift To a Better Bias (Utgoff, 1986) was another early scheme for adjusting bias, but unlike VBMS, STABB was not primarily focussed on searching for bias applicable to large problem domains. Our use of an environment of related tasks in this paper may also be interpreted as an environment of analogous tasks in the sense that ....

Utgoff, P. E. (1986). Shift of bias for inductive concept learning. In Machine Learning: An Artificial Intelligence Approach, pp. 107--147. Morgan Kaufmann.

....the suitable learning operations to apply to a given problem. Eliciting the control of the learning process referred to as declarative bias has always been a fundamental issue in ML because it strongly affects the complexity of the learning process and the learning results [ Mitchell, 1991 ] Utgoff, 1986 ] Grosof and Russell, 1990 ] Three of the works presented at the workshop also demonstrate the utility of a bias language as a concise and powerful way for the user to explicitly parameterize ML systems instead of tuning low level knowledge such as examples and domain theory. As shown by B. ....

P.E. Utgoff. Shift of bias for inductive concept-learning. In R.S Michalski, J.G. Carbonell, and T.M. Mitchell, editors, Machine Learning : an artificial intelligence approach. Morgan Kauffman, 1986.

....in Heskes (1998) 151 BAXTER E Early machine learning work. In Rendell, Seshu, and Tcheng (1987) VBMS or Variable Bias Management System was introduced as a mechanism for selecting amongst different learning algorithms when tackling a new learning problem. STABB or Shift To a Better Bias (Utgoff, 1986) was another early scheme for adjusting bias, but unlike VBMS, STABB was not primarily focussed on searching for bias applicable to large problem domains. Our use of an environment of related tasks in this paper may also be interpreted as an environment of analogous tasks in the sense that ....

Utgoff, P. E. (1986). Shift of bias for inductive concept learning. In Machine Learning: An Artificial Intelligence Approach, pp. 107--147. Morgan Kaufmann.

....by the system automatically is determined to be incorrect, the agent should be able to select a new bias. Solving this problem requires determining when the correct bias is inadequate, finding alternative biases, possibly evaluating competing biases, and finding a theory using the new bias. Utgoff s STABB [1986] is the earliest work that directly addresses this problem. STABB uses version space collapse as a signal that the current bias is inadequate (since no consistent concept can be found) A search is then initiated for a new term to add to the feature hierarchy that captures a necessary distinction. ....

Paul Utgoff. Shift of bias for inductive concept learning. In Ryszard Michalski, Jaime Carbonell, and Tom Mitchell, editors, Machine Learning II, pages 107--148. Morgan Kaufman, 1986.

No context found.

Utgoff, P. E. (1986a). Shift of bias for inductive concept learning. In Michalski, Carbonell & Mitchell (Eds.), Machine learning: An artificial intelligence approach. San Mateo, CA: Morgan Kaufmann.

No context found.

Utgoff, P.: Shift of Bias for Inductive Concept Learning, in: Machine Learning: an artificial intelligence approach (R. Michalski, J. Carbonell, T. Mitchell, Eds.), vol. II, Morgan Kaufmann, Los Altos, CA, 1986, 107--148.

No context found.

P. E. Utgoff. Shift of bias for inductive concept learning. In Michalski, Carbonell, and Mitchell, editors, Machine learning: An artificial intelligence approach, pages 107--148, San Mateo, CA, 1986. Morgan Kaufmann.

No context found.

P. E. Utgoff "Shift of Bias for Inductive Concept Learning" in Machine Learning: An Artificial Intelligence Approach, 1986, Morgan Kaufmann.

No context found.

Utgoff, P. (1986). Shift of bias for inductive concept learning. In R. Michalski et al. (Eds), Machine Learning vol II. Morgan Kaufmann.

No context found.

P.E. Utgoff. Shift of bias for inductive concept learning. In R.S. Michalski, J.G. Carbonell, and T.M. Mitchell, editors, Machine Learning: An AI Approach, Volume II. Morgan Kaufmann, 1986.

No context found.

P. E. Utgoff. Shift of bias for inductive concept learning. In Morgan Kaufmann,ed., Machine learning : An artificial intelligence approach Vol. 2, 1986.

No context found.

Utgoff, P. (1986). Shift of bias for inductive concept learning. In R. Michalski, J. Carbonell and T. Mitchell (Eds.), Machine Learning: An Artificial Intelligence Approach: Vol II (pp. 107-148). Los Altos: Morgan Kaufmann. 10

No context found.

Utgoff, P. (1986). Shift of bias for inductive concept learning. In R. Michalski, J. Carbonell and T. Mitchell (Eds.), Machine Learning: An Artificial Intelligence Approach: Vol II (pp. 107-148). Los Altos: Morgan Kaufmann.

No context found.

P. Utgoff. Shift of bias for inductive concept learning. In R. Michalski, J. Carbonell, and T. Mitchell, editors, Machine Learning, volume 2, pages 163-190. Morgan Kaufmann, Los Altos, CA, 1986.

No context found.

P E Utgoff. Shift of bias for inductive concept learning. In R S Michalski, J G Carbonell, and T M Mitchell, editors, Machine Learning: An Artificial Intelligence Approach (Volume 2). Morgan Kaufmann, 1986.

First 50 documents  Next 50

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