KDD Nuggets 95:29, e-mailed 95-11-10 Contents: * GPS, Journal of Data Mining and Knowledge Discovery -- forthcoming! * K. Thearling, Pilot Software announces data mining initiative, http://www.pilotsw.com * M. Mollo, query: filtering personal information ? Publications: * S. Murthy, PhD thesis: On Growing Better Decision Trees from Data, http://www.cs.jhu.edu/grad/murthy * J. Catlett, Ph.D. thesis: Megainduction: machine learning on very large databases, (1991) online at http://www.research.att.com/orgs/ssr/people/catlett/phd.html Positions: * C. Djeraba, CS position at IRESTE, Nantes, France * I. Haimowitz, GE job opportunities in data mining / warehousing Meetings: * P. Smyth, AISTATS-97: Preliminary Announcement, http://www.stat.washington.edu/aistats97/ * GPS, Rule-Extraction from Neural Networks at AISB-96 -- The KDD Nuggets is a moderated mailing list on Data Mining and Knowledge Discovery in Databases (KDD). Please include a DESCRIPTIVE subject line and a URL, when available, in your submission. Nuggets frequency is approximately weekly. Back issues of Nuggets, a catalog of S*i*ftware (data mining tools), references, and other related information is available at Knowledge Discovery Mine, URL http://info.gte.com/~kdd E-mail add/delete requests to kdd-request@gte.com E-mail contributions with a DESCRIPTIVE subject line to kdd@gte.com. -- Gregory Piatetsky-Shapiro (moderator) ********************* Official disclaimer *********************************** * All opinions expressed herein are those of the writers (or the moderator) * * and not necessarily of their respective employers (or GTE Laboratories) * ***************************************************************************** ~~~~~~~~~~~~ Quotable Quote ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ The road of excess leads to the palace of wisdom. William Blake, (1757-1827) from http://www.mcs.net/~jorn/html/blake.html >~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Date: Mon, 6 Nov 1995 14:33:18 -0500 From: gps@gte.com (Gregory Piatetsky-Shapiro) Subject: New Journal on Data Mining and Knowledge Discovery I am happy to inform the KDD community about a forthcoming new publication: The Journal of Data Mining and Knowledge Discovery which we hope would be the flagship publication in the Data Mining and KDD area and provide a unified forum for KDD research community, whose publications are currently scattered among many different journals. The journal will publish state-of-the-art papers in both research and practice of KDD, surveys of important techniques from related fields, and application papers of general interest. In addition, there will be a section for publishing useful information such as short reports on applications and theory advances (1-3 pages), book and system reviews, and relevant product announcements. The journal will be a quarterly, with a first issue published in January 1997 by Kluwer Academic Publishers. Editors-in-Chief: Usama M. Fayyad ================ Jet Propulsion Laboratory, California Institute of Technology, USA Heikki Mannila University of Helsinki, Finland Gregory Piatetsky-Shapiro GTE Laboratories, USA full announcement and call for papers would be issued soon. >~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Date: Wed, 8 Nov 95 12:03:51 MST From: Kurt Thearling To: kdd@gte.com Subject: Pilot Software announces data mining initiative Content-Type: text Content-Length: 6271 PILOT SOFTWARE LAUNCHES MAJOR NEW DATA MINING INITIATIVE Dun & Bradstreet unit signs prominent business information partners to design large scale, sales and marketing data warehousing applications CAMBRIDGE, Mass. -- November 8, 1995 -- Pilot Software, Inc., a company of The Dun & Bradstreet Corporation and a leading global supplier of advanced business analysis and reporting software, today announced that it has launched a major new initiative focused on creating advanced sales and marketing applications that fully leverage large scale data warehouses using new data mining capabilities that will be delivered as part of LightShip, Pilot's scaleable, on-line analytical processing (OLAP) environment. Pilot has teamed with leading business information providers to offer solutions for the transportation, pharmaceutical and wireless industries. Leading the development effort for Pilot will be a team of scientists, formerly with Thinking Machines Corporation, who are recognized experts in advanced data analysis technology. Data mining techniques allow business users to discover and explore relevant hidden and predictive information housed in massive data warehouses. The automated, prospective analyses offered by data mining move beyond the analyses of past events provided by traditional retrospective data access tools to answer business questions that until now have been too time-consuming to resolve. The data mining tools being developed by Pilot and its partners will enable sales and marketing professionals to predict future trends and behaviors, allowing them to make proactive, knowledge-driven decisions. Users will be able to identify the best customers or prospects for a specific marketing effort, automatically segment those customers by the most relevant attributes and predict the effectiveness of promotional plans. "Pilot today provides the most scaleable OLAP environment in the industry enabling users to quickly implement advanced business analysis applications across their organizations," said John Fleming, Pilot's vice president of marketing. "We are expanding our OLAP capabilities to allow our customers to fully leverage their corporate data warehouses using these new data mining tools. The leading business information providers in credit, pharmaceutical, and wireless are partnering with Pilot to develop state-of-the-art enterprise-class OLAP solutions." Pilot's partners in developing these sales and marketing data mining applications are: Dun & Bradstreet Information Services, a provider of business-to-business information and services, with the world's largest business information database covering over 39 million companies worldwide, is working with Pilot to integrate data mining tools into its Market Spectrum suite of database marketing products and services, with an initial focus on the transportation industry. Using these data mining capabilities, transportation executives will be able to identify potential customers and their purchasing patterns for transportation services from DBIS' data warehouse of millions of businesses worldwide. IMS America, the leading provider of worldwide market research, sales management and decision support services for pharmaceutical and other healthcare related industries, is integrating the data mining tools being developed by Pilot into its Xplorer sales and marketing data warehouse suite of products. With these tools, pharmaceutical companies will be able to manage their product portfolios at a level of detail unattainable before -- allowing them to track sales of multiple products at the prescriber level and allocate marketing resources accordingly. Lightbridge, Inc., the leading provider of integrated customer acquisition and retention services for the wireless industry, is partnering with Pilot to offer a system that provides detailed intelligence for wireless carriers. Lightbridge will integrate its existing customer acquisition solutions for the wireless industry with Pilot's LightShip product to create a unique wireless intelligence system. This system will enable carriers to analyze data and better understand who their customers are, reduce customer acquisition costs and decrease churn. Thus, giving carriers a competitive advantage through a better understanding of their market. "D&B's goal is to create business insight for its customers," said Dennis G. Sisco, executive vice president of The Dun & Bradstreet Corporation. "The powerful data mining and information management initiatives we are announcing today will help us to do that, by extracting exciting new insights from the massive amounts of data now collected both by our customers and by D&B. These initiates will, for the first time, open vast data warehouses to fast, accurate, distributed decision making for our customers." Leading Pilot's development efforts for the data mining initiative is a team of scientists and developers who were responsible at Thinking Machines Corporation for developing a sophisticated suite of data mining tools for parallel supercomputers. Stephen Smith is Chief Scientist for Pilot's Data Intelligence group, which includes Mario Bourgoin, Kurt Thearling, Emily Stone and Joe Yarmus. The team has more than 30 cumulative years' experience in data mining, parallel computation, very large database analysis, machine learning and data visualization. Pilot Software, Inc., a company of The Dun & Bradstreet Corporation, is a leading global supplier of advanced business analysis and reporting software to organizations who need to make critical decisions in rapidly changing environments. Pilot's LightShip product family is the most scaleable OLAP environment for high-performance planning and analysis applications requiring customized multidimensional access and visualization of enterprise information. Pilot Software is headquartered in Cambridge, Mass., and has a strong international presence with offices throughout North and South America, Europe and the Pacific Rim. For more information, Pilot's home page can be reached at http://www.pilotsw.com. ### LightShip is a trademark and Pilot is a registered trademark of Pilot Software, Inc. All other products are trademarks of their respective holders. >~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Return-Path: Date: Thu, 09 Nov 1995 10:02:40 +0100 From: MariaJose.Mollo@CSELT.STET.IT (Mollo Maria Jose`) Subject: filter - filter To: kdd-request@gte.com X-Envelope-To: kdd-request@gte.com Content-Transfer-Encoding: 7BIT Content-Type: text Content-Length: 1713 We are designing a system which makes use of complex filters for the monitoring of very large amounts of data with the objective of finding in them anomalous patterns. These filters are composed of personalized conditions for monitoring data, and can have associated several tables that can be obtained by monitoring the data. Personalization will be the KEY of the system. Each user has different interests that generally change over time. The system should give the users the capability of deciding which filter they want to see and be able to change them as their business needs change. The capability of creating and modifying the filters is a fundamental feature of the system. I would greatly appreciate any suggestions or news that you might have about commercial tools which allow for "filter programmability", so that I can contact the providers as soon as possible. Thank you very much for your kind interest. Best regards Maria Jose` Mollo CSELT (Centro Studi E Laboratori Telecomunicazioni) Via Borgaro, 21 [Sede Via Reiss Romoli, 274] 10148 Torino ITALY Ph. : +39 11 2286958 E_Mail: mollo@CSELT.STET.IT E_Mail: maria.jose.mollo@CSELT.STET.IT >~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Date: Fri, 3 Nov 1995 18:00:37 -0500 (EST) From: Sreerama Murthy Subject: PhD thesis: On Growing Better Decision Trees from Data The following PhD thesis is now available on WWW and FTP. You may retrieve the postscript files for whole thesis as well as just the chapter(s) of interest. There is also a HTML version that allows retrieval of even individual subsections. Title and Abstract: ------------------- On Growing Better Decision Trees from Data Sreerama K. Murthy Department of Computer Science, Johns Hopkins University Thesis Advisor: Steven L. Salzberg This thesis investigates the problem of growing decision trees from data, for the purposes of classification and prediction. After a comprehensive, multi-disciplinary survey of work on decision trees, some algorithmic extensions to existing tree growing methods are considered. The implications of using (1) less greedy search and (2) less restricted splits at tree nodes are systematically studied. Extending the traditional axis-parallel splits to {\it oblique} splits is shown to be practical and beneficial for a variety of problems. However, the use of more extensive search heuristics than the traditional greedy heuristic is argued to be unnecessary, and often harmful. Any effort to build good decision trees from real-world data involves ``massaging'' the data into a suitable form. Two forms of data massaging, domain-independent and domain-specific, are distinguished in this work. A new framework is outlined for the former, and the importance of the latter is illustrated in the context of two new, complex classification problems in astronomy. Highly accurate and small decision tree classifiers are built for both these problems through a collaborative effort with astronomers. To retrieve: ---------------- World-Wide-Web: http://www.cs.jhu.edu/grad/murthy. FTP: Anonymous ftp to blaze.cs.jhu.edu. Directory pub/murthy. The file thesis.ps.gz is the whole thesis. For individual chapters, first get contents.ps.gz, which has the table of contents. The directory contains a postscript file for each chapter. The filenames should be self-explanatory. Dont forget to use binary transfer mode! >~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Date: Fri, 3 Nov 95 19:52 EST From: catlett@research.att.com (Jason Catlett) To: kdd@gte.com Subject: Ph.D. thesis (1991) by Jason Catlett available online Content-Type: text Content-Length: 245 http://www.research.att.com/orgs/ssr/people/catlett/phd.html Various materials are available from my Ph.D. thesis entitled ``Megainduction: machine learning on very large databases'' written in 1991 at the University of Sydney. Jason Catlett >~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ From: cdjeraba@ireste.fr (Chabane DJERABA) Subject: Position To: kdd@gte.com Date: Mon, 6 Nov 1995 08:14:10 +0100 (MET) X-Mailer: ELM [version 2.4 PL24] Content-Type: text Content-Length: 1609 Full professor position in Computer Science at IRESTE, Nantes, France A tenurable vacancy exists for a professor position in Computer Science at the School of Electronic Systems and Computer Engineering (IRESTE). The School offers undergraduate and postgraduate degrees in Computer systems, software and networking. Teaching activities of the school include: - Object models - Databases - Knowledge bases - Networking - Multimedia - Software engineering - Image processing - Human interface Research activities of the school include: - Knowledge Discovery in Databases (extraction algorithms, knowledge and data representation, pre-processing, discovered knowledge evaluations, knowledge discovery in multimedia databases, etc.) - Multimedia Databases (synchronization, retrieval in a multimedia database) - Object-Oriented Databases, interoperability and distributed databases. Qualifications/Skills --------------------- Applicants should possess a doctoral degree in Computer Science/ Information Technology. The appointee will have a strong research profile, will contribute to undergraduate teaching. He will contribute to the development of research activities in one or more areas of computer science developed in the school, and more specially in Knowledge Discovery in Databases, Multimedia databases, Object-Oriented Databases and interoperability. Deadline : February 1996. --------- Prof. Henri BRIAND IRESTE, ATLANPOLE, La Chantrerie, CP 3003 44087 Nantes Cedex 03, FRANCE. Tel. 40 68 30 60, Fax. 40 68 30 66 Tel: (33) 40 68 Fax: (33) 40 68 >~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Date: Wed, 8 Nov 95 08:48:41 EST From: haimowit@oceana.crd.ge.com (Ira Haimowitz) To: kdd@gte.com Subject: GE job opportunities in data mining / data warehousing GE's Corporate Research and Development Center (CRD) has job opportunities for experienced Masters and PhD degree holders in data mining and data warehousing, to participate in large-scale database projects in consumer finance, customer service, and manufacturing quality. ONLY QUALIFIED APPLICANTS AS BELOW, PLEASE. REQUIRED: 1) Hands-on experience in implementing data storage and access solutions for multi-million record databases. OR 2) Hands-on experience in sampling and knowledge discovery analysis algorithms on multi-million record databases; use of commercial software a plus. DESIRABLE: 1) Experience with C++ object-oriented programming See more requirements and details about GE CRD below. If interested, please send resume with cover letter by e-mail or post to: Steve Mirer Information Technology Laboratory GE Corporate Research and development Building K-1, Room 5C11 P.O. Box 8 Schenectady, NY 12301 Internet: mirer@crd.ge.com GE has a variety of openings for Computer Scientists at its Corporate Research and Development Center in Schenectady, New York. GE is one of the world's largest and most successful companies, having leadership positions in business segments including electrical power generation, plastics, medical systems, capital services, aircraft engines, appliances, and lighting. GE's Corporate Research and Development Center (CRD) supports the advanced technology requirements of all GE businesses. The 950 member staff of scientists and engineers is composed of representatives of most major disciplines. CRD is currently offering a variety of employment opportunities for PhDs in Computer Science. Much of this growth arises from opportunities to impact GE's Capital Service, NBC and GE Information Service Businesses. GE is an affirmative action, equal opportunity employer. >~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Date: Tue, 7 Nov 95 09:46:39 PST From: pjs@aig.jpl.nasa.gov (Padhraic J. Smyth) To: CLASS-L@CCVM.SUNYSB.EDU, Connectionists@cs.cmu.edu, ai-stats@watstat.uwaterloo.ca, bayes-news@STAT.CMU.EDU, cogpsy@phil.ruu.nl, dbworld@cs.wisc.edu, ga-list@AIC.NRL.NAVY.MIL, kdd@gte.com, ml@ics.uci.edu, mlnet@swi.psy.uva.nl, news-announce-conferences@uunet.uu.net, nl-kr@snyside1.sunnyside.com, nl-kr@cs.rpi.edu, uai@maillist.CS.ORST.EDU Subject: AISTATS-97: Preliminary Announcement Content-Type: text Content-Length: 1085 Preliminary Announcement Sixth International Workshop on Artificial Intelligence and Statistics (AISTATS-97) January 4-7, 1997 Ft. Lauderdale, Florida This is the sixth in a series of workshops which has brought together researchers in Artificial Intelligence (AI) and in Statistics to discuss problems of mutual interest. The exchange has broadened research in both fields and has strongly encouraged interdisciplinary work. Papers on all aspects of the interface between AI & Statistics are encouraged. For more details consult the AISTATS-97 Web page at: http://www.stat.washington.edu/aistats97/ A full Call for Papers will be released in early 1996. The paper submission deadline will be July 1st 1996. The workshop is organized under the auspices of the Society for Artificial Intelligence and Statistics. Program Chair: David Madigan, University of Washington General Chair: Padhraic Smyth, JPL and UCI >~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Date: Wed, 8 Nov 1995 17:26:46 -0500 From: gps0 (Gregory Piatetsky-Shapiro) To: kdd Subject: AISB-96 Rule-Extraction From Neural Networks workshop FIRST CALL FOR PAPERS AISB-96 WORKSHOP Society for the Study of Artificial Intelligence and Simulation of Behaviour (SSAISB) University of Sussex, Brighton, England April 2, 1996 -------------------------------------------- RULE-EXTRACTION FROM TRAINED NEURAL NETWORKS -------------------------------------------- Robert Andrews Neurocomputing Research Centre Queensland University of Technology Brisbane 4001 Queensland, Australia Phone: +61 7 864-1656 Fax: +61 7 864-1969 E-mail: robert@fit.qut.edu.au Joachim Diederich Neurocomputing Research Centre Queensland University of Technology Brisbane 4001 Queensland, Australia Phone: +61 7 864-2143 Fax: +61 7 864-1801 E-mail: joachim@fit.qut.edu.au Lee Giles NEC Research Institute 4 Independence Way Princeton, NJ 08540 The objective of the workshop is to provide a discussion platform for researchers interested in Artificial Neural Networks (ANNs), Artificial Intelligence (AI) and Cognitive Science. The workshop should be of considerable interest to computer scientists and engineers as well as to cognitive scientists and people interested in ANN applications which require a justification of a classification or inference. INTRODUCTION It is becoming increasingly apparent that without some form of explanation capability, the full potential of trained Artificial Neural Networks may not be realised. The problem is an inherent inability to explain in a comprehensible form, the process by which a given decision or output generated by an ANN has been reached. For Artificial Neural Networks to gain a even wider degree of user acceptance and to enhance their overall utility as learning and generalisation tools, it is highly desirable if not essential that an `explanation' capability becomes an integral part of the functionality of a trained ANN. Such a requirement is mandatory if, for example, the ANN is to be used in what are termed as `safety critical' applications such as airlines and power stations. In these cases it is imperative that a system user be able to validate the output of the Artificial Neural Network under all possible input conditions. Further the system user should be provided with the capability to determine the set of conditions under which an output unit within an ANN is active and when it is not, thereby providing some degree of transparency of the ANN solution. Craven & Shavlik (1994) define the rule-extraction from neural networks task as follows: "Given a trained neural network and the examples used to train it, produce a concise and accurate symbolic description of the network." The following discussion of the importance of rule-extraction algorithms is based on this definition. THE IMPORTANCE OF RULE-EXTRACTION ALGORITHMS Since rule extraction from trained Artificial Neural Networks comes at a cost in terms of resources and additional effort, an early imperative in any discussion is to delineate the reasons why rule extraction is an important, if not mandatory, extension of conventional ANN techniques. The merits of including rule extraction techniques as an adjunct to conventional Artificial Neural Network techniques include: Data exploration and the induction of scientific theories Over time neural networks have proven to be extremely powerful tools for data exploration with the capability to discover previously unknown dependencies and relationships in data sets. As Craven and Shavlik (1994) observe, `a (learning) system may discover salient features in the input data whose importance was not previously recognised.' However, even if a trained Artificial Neural Network has learned interesting and possibly non-linear relationships, these relationships are encoded incomprehensibly as weight vectors within the trained ANN and hence cannot easily serve the generation of scientific theories. Rule-extraction algorithms significantly enhance the capabilities of ANNs to explore data to the benefit of the user. Provision of a `user explanation' capability Experience has shown that an explanation capability is considered to be one of the most important functions provided by symbolic AI systems. In particular, the salutary lesson from the introduction and operation of Knowledge Based systems is that the ability to generate even limited explanations (in terms of being meaningful and coherent) is absolutely crucial for the user-acceptance of such systems. In contrast to symbolic AI systems, Artificial Neural Networks have no explicit declarative knowledge representation. Therefore they have considerable difficulty in generating the required explanation structures. It is becoming increasingly apparent that the absence of an `explanation' capability in ANN systems limits the realisation of the full potential of such systems and it is this precise deficiency that the rule extraction process seeks to redress. Improving the generalisation of ANN solutions Where a limited or unrepresentative data set from the problem domain has been used in the ANN training process, it is difficult to determine when generalisation can fail even with evaluation methods such as cross-validation. By being able to express the knowledge embedded within the trained Artificial Neural Network as a set of symbolic rules, the rule-extraction process may provide an experienced system user with the capability to anticipate or predict a set of circumstances under which generalisation failure can occur. Alternatively the system user may be able to use the extracted rules to identify regions in input space which are not represented sufficiently in the existing ANN training set data and to supplement the data set accordingly. A CLASSIFICATION SCHEME FOR RULE EXTRACTION ALGORITHMS The method of classification proposed here is in terms of: (a) the expressive power of the extracted rules; (b) the `translucency' of the view taken within the rule extraction technique of the underlying Artificial Neural Network units; (c) the extent to which the underlying ANN incorporates specialised training regimes; (d) the `quality' of the extracted rules; and (e) the algorithmic `complexity' of the rule extraction/rule refinement technique. The `translucency' dimension of classification is of particular interest. It is designed to reveal the relationship between the extracted rules and the internal architecture of the trained ANN. It comprises two basic categories of rule extraction techniques viz `decompositional' and `pedagogical' and a third - labelled as `eclectic' - which combines elements of the two basic categories. The distinguishing characteristic of the `decompositional' approach is that the focus is on extracting rules at the level of individual (hidden and output) units within the trained Artificial Neural Network. Hence the `view' of the underlying trained Artificial Neural Network is one of `transparency'. The translucency dimension - `pedagogical' is given to those rule extraction techniques which treat the trained ANN as a `black box' ie the view of the underlying trained Artificial Neural Network is `opaque'. The core idea in the `pedagogical' approach is to `view rule extraction as a learning task where the target concept is the function computed by the network and the input features are simply the network's input features'. Hence the `pedagogical' techniques aim to extract rules that map inputs directly into outputs. Where such techniques are used in conjunction with a symbolic learning algorithm, the basic motif is to use the trained Artificial Neural Network to generate examples for the learning algorithm. As indicated above the proposed third category in this classification scheme are composites which incorporate elements of both the `decompositional' and `pedagogical' (or `black-box') rule extraction techniques. This is the `eclectic' group. Membership in this category is assigned to techniques which utilise knowledge about the internal architecture and/or weight vectors in the trained Artificial Neural Network to complement a symbolic learning algorithm. An ancillary problem to that of rule extraction from trained ANNs is that of using the ANN for the `refinement' of existing rules within symbolic knowledge bases. The goal in rule refinement is to use a combination of ANN learning and rule extraction techniques to produce a `better' (ie a `refined') set of symbolic rules which can then be applied back in the original problem domain. In the rule refinement process, the initial rule base (ie what may be termed `prior knowledge') is inserted into an ANN by programming some of the weights. The rule refinement process then proceeds in the same way as normal rule extraction viz (1) train the network on the available data set(s); and (2) extract (in this case the `refined') rules - with the proviso that the rule refinement process may involve a number of iterations of the training phase rather than a single pass. DISCUSSION POINTS FOR WORKSHOP PARTICIPANTS 1. Decompositional vs. learning approaches to rule- extraction from ANNs - What are the advantages and disadvantages w.r.t. performance, solution time, computational complexity, problem domain etc. Are decompositional approaches always dependent on a certain ANN architecture? 2. Rule-extraction from trained neural networks vs. symbolic induction. What are the relative strength and weaknesses? 3. What are the most important criteria for rule quality? 4. What are the most suitable representation languages for extracted rules? How does the extraction problem vary across different languages? 5. What is the relationship between rule-initialisation (insertion) and rule-extraction? For instance, are these equivalent or complementary processes? How important is rule-refinement by neural networks? 6. Rule-extraction from trained neural networks and computational learning theory. Is generating a minimal rule-set which mimics an ANN a hard problem? 7. Does rule-initialisation result in faster learning and improved generalisation? 8. To what extent are existing extraction algorithms limited in their applicability? How can these limitations be addressed? 9. Are there any interesting rule-extraction success stories? That is, problem domains in which the application of rule-extraction methods has resulted in an interesting or significant advance. ACKNOWLEDGEMENT Many thanks to Mark Craven, and Alan Tickle for comments on earlier versions of this proposal. RELEVANT PUBLICATIONS Andrews, R Diederich, J and Tickle, A.B.: A survey and critique of techniques for extracting rules from trained artificial neural networks. To appear: Knowledge-Based Systems, 1995 (ftp:fit.qut.edu.au//pub/NRC/ps/QUTNRC-95-01- 02.ps.Z) Andrews, R and Geva, S: `Rule extraction from a constrained error back propagation MLP' Proc. 5th Australian Conference on Neural Networks Brisbane Queensland (1994) pp 9-12 Andrews, R and Geva, S `Inserting and extracting knowledge from constrained error back propagation networks' Proc. 6th Australian Conference on Neural Networks Sydney NSW (1995) Craven, M W and Shavlik , J W `Using sampling and queries to extract rules from trained neural networks' Machine Learning: Proceedings of the Eleventh International Conference (San Francisco CA) (1994) (in print) Diederich, J `Explanation and artificial neural networks' International Journal of Man-Machine Studies Vol 37 (1992) pp 335-357 Fu, L M `Neural networks in computer intelligence' McGraw Hill (New York) (1994) Fu, L M `Rule generation from neural networks' IEEE Transactions on Systems, Man, and Cybernetics Vol 28 No 8 (1994) pp 1114-1124 Gallant, S `Connectionist expert systems' Communications of the ACM Vol 31 No 2 (February 1988) pp 152-169 Giles, C L and Omlin C W `Rule refinement with recurrent neural networks' Proc. of the IEEE International Conference on Neural Networks (San Francisco CA) (March 1993) pp 801-806 Giles, C L and Omlin C W `Extraction, insertion, and refinement of symbolic rules in dynamically driven recurrent networks' Connection Science Vol 5 Nos 3 and 4 (1993) pp 307-328 Giles, C L, Miller, C B, Chen, D, Chen, H, Sun, G Z and Lee, Y C `Learning and extracting finite state automata with second-order recurrent neural networks' Neural Computation Vol 4 (1992) pp 393-405 Hayward, R.; Pop, E.; Diederich, J.: Extracting Rules for Grammar Recognition from Cascade-2 Networks. Proceeding, IJCAI-95 Workshop on Machine Learning and Natural Language Processing. McMillan, C, Mozer, M C and Smolensky, P `The connectionist scientist game: rule extraction and refinement in a neural network' Proc. of the Thirteenth Annual Conference of the Cognitive Science Society (Hillsdale NJ) 1991 Omlin, C W, Giles, C L and Miller, C B `Heuristics for the extraction of rules from discrete time recurrent neural networks' Proc. of the International Joint Conference on Neural Networks (IJCNN'92) (Baltimore MD) Vol 1 (1992) pp 33 Pop, E, Hayward, R, and Diederich, J `RULENEG: extracting rules from a trained ANN by stepwise negation' QUT NRC (December 1994) Sestito, S and Dillon, T `Automated knowledge acquisition of rules with continuously valued attributes' Proc. 12th International Conference on Expert Systems and their Applications (AVIGNON'92) (Avignon France) (May 1992) pp 645-656. Sestito, S and Dillon, T `Automated knowledge acquisition' Prentice Hall (Australia) (1994) Thrun, S B `Extracting Provably Correct Rules From Artificial Neural Networks' Technical Report IAI-TR-93-5 Institut fur Informatik III Universitat Bonn (1994) Tickle, A B, Orlowski, M, and Diederich, J `DEDEC: decision detection by rule extraction from neural networks' QUT NRC (September 1994) Towell, G and Shavlik, J `The Extraction of Refined Rules Tresp, V, Hollatz, J and Ahmad, S `Network Structuring and Training Using Rule-based Knowledge' Advances In Neural Information Processing Vol 5 (1993) pp871-878 SUBMISSION OF WORKSHOP EXTENDED ABSTRACTS/PAPERS Authors are invited to submit 3 copies of either an extended abstract or full paper relating to one of the topic areas listed above. Papers should be written in English in single column format and should be limited to no more than eight, (8) sides of A4 paper including figures and references. Centered at the top of the first page should be complete title, author name(s), affiliation(s), and mailing and email address(es), followed by blank space, abstract(15-20 lines), and text. Please include the following information in an accompanying cover letter: Full title of paper, presenting author's name, address, and telephone and fax numbers, authors e-mail address. Submission Deadline is January 15th,1996 with notification to authors by 31st January,1996. For further information, inquiries, and paper submissions please contact: Robert Andrews Queensland University of Technology GPO Box 2434 Brisbane Q. 4001. Australia. phone +61 7 864-1656 fax +61 7 864-1969 email robert@fit.qut.edu.au More information about the AISB-96 workshop series is available from: ftp: ftp.cogs.susx.ac.uk pub/aisb/aisb96 WWW: (http://www.cogs.susx.ac.uk/aisb/aisb96) WORKSHOP PARTICIPATION CHARGES The workshop fees are listed below. Note that these fees include lunch. Student charges are shown in brackets. AISB NON-ASIB MEMBERS MEMBERS 1 Day Workshop 65 (45) 80 LATE REGISTRATION: 85 (60) 100 PROGRAM COMMITTEE MEMBERS R. Andrews, Queensland University of Technology A. Tickle, Queensland University of Technology S. Sestito, DSTO, Australia J. Shavlik, University of Wisconsin >~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~