HARUHISA TAKAHASHI

1. Nonlinear networks :Synthesis and analysis of nonlinear electrical networks; Several important concepts on linear theory had been generalized as linear concepts.
2. Queueing theory: Especially the diffusion approximation of queueing system had been proposed including two-dimensional jump diffusion process which was analytically solved by showing Fokker-Planc equation and boundary value problem.
3. Performance evaluation of communication systems: applying the queueing analysis several communication system was analized.
4. Neural networks: Artificial neural networks are studied, giving learning algorithms and clarifying the network ability theoretically.
5. Learning theory: Generalization ability was analized by proposing a new technique which can give better estimation than VC dimension analysis.
6. Learning algorithms: Improvements of the Support vector machin algorithm was given.
7. Mean field approximation of the graphical models: A new approximatio technique of graphical models are studied.