Information theoretic competitive learning in self-adaptive multi-layered networks

In this paper, we propose self-adaptive multi-layered networks in which information in each processing layer is always maximized. Using these multi-layered networks, we can solve complex problems and discover salient features that single-layered networks fail to extract. In addition, this successive information maximization enables networks gradually to extract important features. We applied the new method to the Iris data problem, the vertical-horizontal lines detection problem, a phonological data analysis problem and a medical data problem. Experimental results confirmed that information can repeatedly be maximized in multi-layered networks and that the networks can extract features that cannot be detected by single-layered networks. In addition, features extracted in successive layers are cumulatively combined to detect more macroscopic features.     

Information enhancement for interpreting competitive learning

In this paper, we propose a new method called information enhancement to interpret internal representations of competitive learning. We consider competitive learning as a process of mutual information maximisation on input patterns. Then, we examine to what extent this mutual information can be increased or decreased by focusing upon or enhancing some elements in a network. If this enhancement for the elements increases information on input patterns, these elements possess more information on input patterns. Thus, we only have to carefully examine those elements in a network. We applied the method to an artificial problem, the Iris problem and an air pollution problem. In all problems, we succeeded in extracting important features in patterns. In addition, final maps were better than those obtained by the conventional self-organising map. We can say that this is the first step towards the full understanding of internal representations in competitive learning.     

Optical property of poly(9,9-dioctylfluorene) gel with β phase and application to polymer light emitting diode

Poly (9,9-dioctylfluorene)(F8) gel with β phase has been investigated in terms of optical absorption, photoluminescence measurements and Fourier transform infrared spectroscopy. The optical properties of the F8 gel markedly changed in the temperature range from 70 to 80 °C owing to the vibration of polymer chain related to the glass transition temperature of F8. F8 films by thermal printing method had the characteristic particulate morphology. Current efficiencies of polymer light emitting diodes (PLEDs) with the β phase of F8 fabricated by the thermal printing method were better than that with the amorphous phase of F8 by the spin-coating method. We demonstrate the β phase effects of PLEDs characteristics by the thermal printing method.     

The numerical analysis of particle-size distribution of clusters in shear flow at one-dimensional closed system and three-dimensional open system

Aimed at optimizing the resin-molding process, a method for numerically analyzing aggregation and dispersion behavior of the filler in resin composite was proposed. The flow of a resin composite during molding was calculated by using computational fluid dynamics (CFD), and particle-size distribution (PSD) of the cluster in each computational domain of the CFD model was estimated by solving the population balance equation (PBE). The proposed numerical-analysis method is based on the thixotropy model of Usui et al. In the thixotropy model, PSD is calculated by taking into account the aggregation and dispersion rate of the cluster by Brownian coagulation, shear coagulation, and shear breakage. Shear-breakage rate of the cluster is evaluated by solving the energy balance of bonding energy of the primary particles cut at the breakage and the drag applied to the cluster by the flow of the fluid. The composite viscosity was calculated using Krieger and Dougherty’s model based on apparent-solid-volume fraction estimated from the calculated PSD. To solve the PBE at low calculation cost, it was discretized using the fixed-pivot technique of Kumar and Ramkrishna. The proposed method was incorporated into the general-purpose CFD software FLOW-3D®, and its accuracy was proved.