Online virtual laboratory is important for online science education. CyberLab is a toolkit for online virtual experiments. It is able to handle the creation, exportation, and execution of virtual experiments. We believe CyberLab is the first toolkit that allows instructors, who do not know computer programming, to create their own online virtual experiments. This paper briefly introduces the underlying model and the architecture of CyberLab.

Overfitting is a fundamental problem of most machine learning techniques, including genetic programming (GP). Canary functions have been introduced in the literature as a concept for preventing overfitting by automatically recognizing when it starts to occur. This paper presents a simple scheme for implementing canary functions using cross-validation. The effectiveness of this technique is demonstrated by applying it to the numeric regression problem. A list of conditions and criteria for applying this technique to other problem domains is also identified. Other strategies for dealing with overfitting in GP are discussed.

Despite their strengths, object-oriented methods (OOMs) and their supporting CASE tools often do not produce model that are amenable to rigorous semantic analysis. This is a direct result of their loosely defined semantics. The authors outline their ongoing work on providing a semantic base for OOMs.

A primary motivation for the development of the Connection Machine (CM) was to create a vehicle for artificial intelligence research. The original design was largely based upon Fahlman's NETL machine [Fah79], the primary purpose of which was to effect large semantic networks, a paradigm of artificial intelligence. To date, however, only a small amount of AI research is being conducted on the CM. Discounting neural net and computer vision research, the amount is miniscule. The lack of AI tools for the CM is a primary cause for this dearth of research. AI researchers proposing to use the CM must first develop the necessary AI tools before beginning work on their projects. For example, there are no existing inference systems, knowledge representation packages, expert system toolkits, etc. PARKA, a pseudo-acronym for “Parallel Knowledge Representation and Association”, was developed as the prototype of one such tool: a knowledge representation system modeled on a frame-based representation language (FDL) [Tou87] paradigm.