Phasetransition-like Changes in Human Visual Information Processing

Péter Nagy (Kecskemét College Faculty of Mechanical Engineering and Automation, Hungary)

Istvan Pintér (Kecskemét College Faculty of Mechanical Engineering and Automation, Hungary)

Mihály Bagány (Kecskemét College Faculty of Mechanical Engineering and Automation, Hungary)

Abstract: For exact examination of human information processing ability we elaborated an information-theory-based model and a new measuring method. Starting from the theoretical background of the well-known Hick-Hyman law and analysing the data acquired during the Soyuz-Salyut space missions, an important fact was derived. When examining the model by reducing the H(X) input entropy of the stimulus signals (symbols) and approaching to the 0 bit there was an interesting effect: through the reduction of H(X) input entropy towards zero the ratio of (processed information)/(input entropy) was increasing and its value approached to 1, while at non-zero entropy this ratio was less than 1. Therefore a phase-change-like effect occurred below H(X) = 1 bit. Consequently the zero bit measurement could not be involved into the determination of Information Processing Ability (IPA) while the traditional measurements were performed only at two distinct values of entropy: at 0 bit and at a not too high H(X) value (e.g. 2 bits/symbol).

Keywords: artificial neural networks, associative memory, information processing ability, phase transition, visual information processing

Categories: H.1.0, H.1.1, J.2, J.3, J.7