Basic Technologies for Knowledge Transfer in Intelligent Systems

Abstract

Knowledge transfer is one of the most important mechanisms of human evolution. The ontogeny of humans enables them to act efficiently in a very dynamic environment. Thus, it would be highly desirable to enable "intelligent" artificial systems to behave in a similar way. This article introduces basic technologies that are needed for that purpose. With these technologies -components of a future knowledge transfer toolboxit is possible to detect novel concepts that arise in the input space of a classifier or existing classification rules that become obsolete. Then, prototypes of new rules can be created automatically using an on-line clustering mechanism. These prototypes are compared to already existing rules, rated, and eventually accepted or discarded. In case of acceptance, a human expert labels the rules which are then both integrated into the "own" classifier and sent to other classifiers. Thus, knowledge transfer between "intelligent" artificial systems becomes possible and the overall system is provided with a new kind of self-optimization ability.

Knowledge Transfer Between Artificial Intelligence Systems

“…Knowledge transfer between Artificial Intelligent systems has been the subject of extensive discussion in the literature for more than two decades (Gilev et al, 1991 ; Jacobs et al, 1991 ; Pratt, 1992 ; Schultz and Rivest, 2000 ; Buchtala and Sick, 2007 ) (see also a comprehensive review Pan and Yang, 2010 ). Several technical ideas to achieve AI knowledge transfer have been explored to date.…”

On the versatility of radial basis function neural networks: A case study in the field of intrusion detection

“…It is also evident that DT cannot be built such that they exhibit generative properties as they are not based on local models of the data. In [77,78], we have shown that an RBF can be defined in a way such that it is functionally equivalent to a Mamdani-type FS. For this slightly modified paradigm-which we call RBFS (radial basis function fuzzy system)-the definition of RBF must be changed as follows: The activation of each hidden neuron j 2 U H is determined using a hyperelliptical Gaussian basis function: can be regarded as a center of a hyperellipsoidal cluster in the input space and r ðHÞ j defines the shape of the cluster.…”

“…As this factor is computed without knowledge of the correct label, this experiment demonstrates the potential of the mechanism for real scenarios. In the case of the detection of a novel process, it is possible to react by informing a human domain expert or by generating a new rule prototype that must be labeled by a human domain expert, e.g., a system operator (as we suggest in [78]). It must be emphasized that the example above is quite simplified and ''artificial" in order to visualize it: Usually, more than two input features are used, but the principle of novelty detection is the same.…”

Learning by teaching versus learning by doing: Knowledge exchange in organic agent systems

“…It should be mentioned that this architecture resembles existing approaches for the design of adaptive systems such as the approaches described in [8], [22], for instance. Some more details on this architecture and on specific components can be found in [3], [11]. In this article, we describe some new techniques and measures for the components that deal with knowledge (rule) representation, acquisition, and assessment in more detail.…”

“…For that purpose, we will measure the advantages that may arise from a knowledge exchange, the costs for integrating a teacher into the learning process, and the communication costs. The knowledge acquisition, exchange, and fusion mechanisms are partly based on own, earlier work on this topic (see, e.g., [3], [11], [12], [13], [14], [15] for some more details).…”