The ability of learning networks to generalize can be greatly enhanced by providing constraints from the task domain. This paper demonstrates how such constraints can be integrated into a backpropagation network through the architecture of the network. This approach has been successfully applied to the recognition of handwritten zip code digits provided by the U.S. Postal Service. A single network learns the entire recognition operation, going from the normalized image of the character to the final classification.

Resistance fluctations in submicrometer narrow Si inversion layers are studied over a wide range of temperatures and electron concentrations. Thermally activated switching on and off of discrete resistance increments is observed, caused by the capture and emission of individual electrons at strategically located scatterers (interface traps). The traps have a broad distribution of activation energies, as assumed in accounting for 1/f noise in larger devices.

No abstract

No abstract

No abstract