In the last few years there has been growing interest on amalgamated programming languages which have integrated the features of logic and functional programming languages. The programmer can write programs more naturally in the amalgamated language. But its execution mechanism generally involves a very large search space and is inefficient. Thus interpreters of amalgamated languages make sacrifices in the scope of the language and/or efficiency of the interpreter. In this paper, we propose an amalgamated language, AMLOG, which is equipped with two kinds of equations, '=' and ':='. These symbols are used to control the execution strategy and switch between non-deterministic execution and deterministic (pruned computation tree) execution. Thus, in AMLOG, a programmer can select an appropriate description for a problem. This enlarges the scope of the interpreter and at the same time allows efficient execution.

This paper describes the implementation of a concurrent search algorithm on the distributed system which consists of 15 processors connected by VME-bus. In such a system, it is necessary to exchange messages between asynchronous processors. The efficiency of the system goes down when message exchanges are quite often. An effort is required to make the message exchanges as less as possible. The requirment would be reflected in the problem decomposition methods, namely, to decompose a problem into mutually independent subproblems. We proposed two distinct methods answering the requirment. In this paper, we apply one of them, called Hypothesis Verification Method, and describe the realization of a distributed system appropriate for it. We report on the design and evaluation. Especially, the following two points are investigated : (1) How fast can the system solve 2×2×2 Rubik's Cube ; (2) There exist several problems for which the speed-up ratio superlinearly increases with the number of processors. This phenomenon is called extraordinary acceleration. We investigate the reason why this phenomenon occurs.

We suggest the method to generate the instructional knowledge with the instructional knowledge acquired by the learning system. Our method can classify the given problems, decide the order in which the problem groups should be taught, and generate the easily understandable text. At first, a teacher gives the learning system included in the ICAI system the input consisting of basic operators, problems, worked examples. The basic operators are primitive operators for solving problems, and the worked examples are sequences consisting of applied operators in the problem solving. From these input, the leaning system can obtain the strategy knowledge for controlling the applications of operators. The strategy knowledge are macro-operators and heuristic operators. The classification of problems are done by the analysis of worked examples. We assume that the problems classified into the same group should need the same strategy knowledge. The strategy knowledge are represented by the set of operators applied in the worked example. Therefore, we can classify the problems by the applied operators. Next, the order for teaching the problem groups is determined. Since we consider that the easy problems should be taught at first, the teaching sequence is determined in the order of the smaller strategy knowledge. Furthermore, the layered worked examples are generated by the expansion of applied macro-operators. Finally, from all gained instructional knowledge, the ICAI system can generate the text with the worked examples which students can easily understand.

The relationship between higher order intuitionistic many sorted predicate logic (denoted by L) and the type systemλPK2 (Coquand-Huet's Calculus of Constructions) is investigated using the formulae-as-types notion. As a necessary consequence, the type system λKF which is a restricted version ofλPK2 is obtained to interpret just L. The λKF has made it possible to arrive at an inverse interpretation to L. The interpretation is more constructive than the existing one in the sense that it may also be applied to a proof figure to get a judgment constructively. The soundness and completeness results are given. In this sense, the whole logical system can be translated into the type system, and hence proofs in L are treated formally in λKF.