SPECIAL SECTION UPDATE (Correction~Update of Machine Learning Special Section Sigart 76) MACHINE LEARNING SUMMARY lCarbonelll RESEARCH Personnel: Jaime G. Carbonell and several graduate students (currently Peter Neuss, Gregory Hood and Monica Lam). Carnegie-Mellon University, Computer Science Department. the model consists of a reminding phase (inspired by Schank's MOPS theory) followed by an analogical mapping phase. The latter process extends Means-Ends Analysis to apply in the solutiontransformation space; operators transform recalled solutions to similar problems into a solution satisfying the requirements of the new problem. The learning component of this system consists of organizing new experiences so they will be recalled in appropriate future circumstances, generalizing solutions to normative-plan schemas, acquiring new operators that apply in the analogy transform space, and inducing heuristics for the utility of different analogical mappings, as a function of problem characteristics and areas of the problem solver's expertise. Automated generation of new problems is an integral process to verify the validity of any newly-generalized plan schema or selection heuristic. (At present, the problem generation process is focused only on generating potentially interesting variants of previously-solved problems.) Objectives, methods, and results: THE META PROJECT META is a Natural Language learning project whose specific objective is to acquire new word definitions and new concepts from contextual information in interactive dialogues. It is an instance of a learning-from-examples method, with a difference: learning proceeds in a reactive, knowledge-rich environment. Our initial research indicates that the interactive nature of the environment ought to be a crucial component of any general learning system. In brief, our investigations have led us to formulate the following hypotheses, which we intend to test and pursue further in the immediate future: 1. Learning requires progressive refinement -- It is unreasonable to expect that a computer system (or a human) learn a concept or a skill without error, in its full embellished form, in one brief learning session. It must undergo a sequence of progressive test-andIn other words, a concept can be update stages. learned by first inducing a rough approximation of its and successively correcting this final form approximation with more accurate, more detailed information. 2. Interaction with a reactive environment -- Interaction is the engine that drives learning processes. A learner must be able to direct queries to its teacher or perform experiments on its environment. It must be able test out new concepts and skills as it learns them. 3. Reasoning by Analogy -- The more a system can learn by relating new concepts to old, by modifying existing concepts, or using chunks of existing concepts as building blocks, the more robust and general its learning mechanisms will be. META is a vehicle for investigating these hypotheses. We are currently in the process of developing parts of the computer implementation. LEARNING AND PROBLEM SOLVING BY ANALOGY Learning does not occur in the absence of other cognitive demands. This project is focused in part on viewing learning as an essential component of understanding, problem solving and memory organization, and in part on the need to extend an interrelated memory model -- rather than simply acquiring a self-contained, disembodied "concept". Humans improve their problem-solving skills in given domains simply by virtue of practice solving similar problems in the same domains. We developed a model of problem solving by analogy that exploits prior problem-solving experience. Essentially Support: Our primary funding source is the Office of Naval Research (Grant number N00014-79-C-0661), which is also supporting other Learning research in Psychology and Computer Science at CMU. ONR also supported the CMU workshop-symposium on Machine Learning last summer. We gratefully acknowledge their generous support.

/lp/association-for-computing-machinery/machine-learning-research-u2YJDrPXQL