LiMin Fu

YNIMG Journal 1996 Journal Article

Using self-organizing maps for abstraction of FMRI images

• Didem Gokcay
• LiMin Fu

AIIM Journal 1994 Journal Article

Polygenic trait analysis by neural network learning

• LiMin Fu

AI techniques have been applied to the domain of DNA sequence analysis in predicting or identifying certain specialized regions, in recognizing genes, and in understanding the evolutionary relationships between sequences. This paper focuses on a kind of genetic pattern recognition, namely, the problem of identifying the gene combinations (patterns) causally related to a given trait determined by multiple genes (a so-called polygenic trait). A novel approach is presented which combines neural-network and knowledge-based techniques. The neural network is trained to predict the trait and then the knowledge embedded in the network is decoded into symbolic patterns. This hybrid approach is evaluated in the domain of identifying genes of insulin dependent diabetes mellitus. The consistency between the results with this approach and those reported in genetic literature supports the viability of this approach.

AAAI Conference 1991 Conference Paper

Rule Learning by Searching on Adapted Nets

• LiMin Fu

If the backpropagation network can produce an inference structure with high and robust performance, then it is sensible to extract rules from it. The KT algorithm is a novel algorithm for generating rules from an adapted net efficiently. The al ff orithm is able to deal with both single-layer and mu ti-layer networks, and can learn both confirming and disconfirming rules. Empirically, the algorithm is demonstrated in the domain of wind shear detection by infrared sensors with success.