Provost, F.J. and Kolluri, V., A Survey of Methods for Scaling Up Inductive Learning Algorithms, Proc. 3rd International Conference on Knowledge Discovery and Data Mining, 1997.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....learning algorithm won t be able to see all data in one single batch, and is not allowed to remember too much of the data scanned in the past. As a consequence, scaling up classical learning algorithms to handle extremely large data sets and meet these requirements is an important research issue [12], 2] One approach to satisfy these constraints is to consider incremental learning techniques, in which only a subset of the data is to be considered at each step of the learning process. The learner is therefore incrementally trained as new batches of data are loaded into memory. Support ....

F. J. Provost and V. Kolluri, "A survey of methods for scaling up inductive learning algorithms ", Technical Report ISL-97-3, Intelligent Systems Lab., Department of Computer Science, University of Pittsburgh, 1997.

....itself get computationally intractable. This issue is becoming more evident today, because there are complex classi cation problems waiting to be solved in many domains, where large amounts of training data are already available [9] Researchers in the machine learning and data mining community [6, 5, 11] have therefore been trying to scale up classical inductive learning algorithms to handle extremely large data sets. One popular approach for dealing with the intractability problem of learning from huge databases is to select small data subsets for training. Training with a small data sample ....

F. J. Provost and V. Kolluri. A survey of methods for scaling up inductive learning algorithms. Technical Report ISL-973, Intelligent Systems Lab., Department of Computer Science, University of Pittsburgh, 1997.

....jf (x) yjdP (x; y) When developing classi ers using learning methods, more training data can reduce risk, but the learning process can get computationally intractable. This issue is becoming more evident with large amounts of data available [12] Researchers in machine learning and data mining [7, 8, 22] have therefore been trying to scale up classical inductive learning algorithms to handle extremely large data sets. Scalability is the next big challenge being tackled by the Knowledge Discovery and Data Mining (KDD) community [11, 13, 21] Ideally, it is desirable to consider all the examples ....

F. J. Provost and V. Kolluri. A survey of methods for scaling up inductive learning algorithms. Technical Report ISL-97-3, Intelligent Systems Lab., Department of Computer Science, University of Pittsburgh, 1997.

.... cation problems waiting to be solved in many domains (e.g. scienti c elds, sales logs, social economic nancial studies) where large amounts of data are already available [ Fayyad et al. 1996 ] Researchers in the machine learning and data mining community [ Catlett, 1991a; 1991b; Provost and Kolluri, 1997 ] have therefore been trying to scale up classical inductive learning algorithms to handle extremely large data sets. Ideally, it is desirable to be able to consider all the examples simultaneously, to get the best possible estimate of class distribution. On the other hand when the training ....

F. J. Provost and V. Kolluri. A survey of methods for scaling up inductive learning algorithms. Technical Report ISL-97-3, Intelligent Systems Lab., Department of Computer Science, University of Pittsburgh, 1997.

....1 Introduction When developing classi ers using learning methods, more training data can reduce risk, but the learning process can get computationally intractable. This issue is becoming more evident with large amounts of data available [5] Researchers in machine learning and data mining [2, 3, 7] have therefore been trying to scale up classical inductive learning algorithms to handle extremely large data sets. Ideally, it is desirable to consider all the examples together for the best learning performance. However, when the training set is large, not all data can be loaded into the ....

F. J. Provost and V. Kolluri. A survey of methods for scaling up inductive learning algorithms. Technical Report ISL-973, Intelligent Systems Lab., Department of Computer Science, University of Pittsburgh, 1997.

....This process may be continued iteratively resulting in a hierarchy of meta classifiers. A number of different learning algorithms are available through this system. Another approach to multi agent based distributed machine learning is described in (Provost Hennessy, 1996; Provost Aronis, 1996; Provost Venkateswarlu, 1998). The PADMA system (Kargupta, Hamzaoglu, Stafford, Hanagandi, Buescher, 1996; Kargupta, Hamzaoglu, Stafford, 1997) achieves scalability by locating agents with the distributed data sources. An agent coordinating facilitator gives user requests to local agents which then access and analyze ....

....set of theories results. He then casts this idea in terms of partial reasoning which he then relates to knowledge discovery. A basic requirement of algorithms employed in DDM is that they have the ability to scale up. A survey of methods of scaling up inductive learning algorithms is presented in (Provost Venkateswarlu, 1998). An important class of DDM problems often encountered in many practical DDM applications is that of vertically partitioned feature spaces. For the most part, DDM research has not adequately addressed the common occurrence of heterogeneous, distributed, data sets. An example of mining from ....

Provost, F., & Venkateswarlu, K. (1998). A survey of methods for scaling up inductive learning algorithms.

....out of concept) that are then propagated through databases looking for values where in or out of concept markers accumulate. A basic requirement of algorithms employed in DDM is that they have the ability to scale up. A survey of methods of scaling up inductive learning algorithms is presented in (Provost Venkateswarlu, 1998). 2.3 Overview of Multivariate Regression Multivariate regression (MR) is a widely used data analysis technique owing to its ease of use and intuitive theoretical basis (Mosteller Tukey, 1977; Flury Riedwyl, 1988) MR involves fitting a parametric function model to a set of data. In this ....

Provost, F., & Venkateswarlu, K. (1998). A survey of methods for scaling up inductive learning algorithms.

....We believe it is possible to provide these services in a way which is to some extent generic, in the sense that it is independent of exactly which algorithm or tool is being implemented. Most of the existing literature, including much of the work on scalable knowledge discovery algorithms [6], assumes that the database to be explored is stored in local main memory. But it is quite common to find a client accessing a serverresident database over a network, and the efficiency of algorithms as reported in published papers often depends on programming techniques that cannot be ported to ....

Provost, F.J., Kolluri, V.: A Survey of Methods for Scaling Up Inductive Learning Algorithms. Proc. 3rd International Conference on Knowledge Discovery and Data Mining (1997)

....based on this high level view. We regret that because of space limitations, it is impossible in this paper to provide references to even a fraction of the relevant published work. Instead, we have made available a detailed survey and comprehensive bibliography as a technical report (Provost Kolluri 1997). Copyright 1997, American Association for Artificial Intelligence (www.aaai.org) All rights reserved. Why scale up The most commonly cited reason for attempting to scale inductive methods up to massive data sets is based on the prevailing view of data mining as classifier learning. When ....

Provost, F.J., and Kolluri, V. 1997. A Survey of Methods for Scaling Up Inductive Learning Algorithms, Technical Report ISL-97-3, Intelligent Systems Lab., University of Pittsburgh (http://www.pitt.edu/~uxkst/survey-paper.ps).

No context found.

Provost, F.J. and Kolluri, V., A Survey of Methods for Scaling Up Inductive Learning Algorithms, Proc. 3rd International Conference on Knowledge Discovery and Data Mining, 1997.

No context found.

F. J. Provost and V. Kolluri, "A survey of methods for scaling up inductive learning algorithms", Technical Report ISL-97-3, Intelligent Systems Lab., Department of Computer Science, University of Pittsburgh, 1997.

No context found.

F.J. Provost and V. Kolluri. A survey of methods for scaling up inductive learning algorithms. To appear in: Data Mining and Knowledge Discovery journal. 1998.

No context found.

Provost, F.J. and Kolluri, V. 1997. A Survey of Methods for Scaling Up Inductive Learning Algorithms.

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