Distributed memory systems for simulating artificial neural networks

In executing tasks involving intelligent information processing, the human brain performs better than the digital computer. The human brain derives its power from a large number [O(10¹¹)] of neurons which are interconnected by a dense interconnection network [O(10⁵) connections per neuron]. Artificial neural network (ANN) paradigms adopt the structure of the brain to try to emulate the intelligent information processing methods of the brain. ANN techniques are being employed to solve problems in areas such as pattern recognition, and robotic processing. Simulation of ANNs involves implementation of large number of neurons and a massive interconnection network. In this paper, we discuss various simulation models of ANNs and their implementation on distributed memory systems. Our investigations reveal that communication-efficient networks of distributed memory systems perform better than other topologies in implementing ANNs.     

Optimal systems for moving-object detection and classification

Translated from Kibernetika i Sistemnyi Analiz, No. 5, pp. 137–145, September–October, 1993.     

Method for error recovery of line spectral frequencies

The authors have exploited the ideas used in vector quantisation for error recovery of scalar quantised LSFs. The good performance of this method has provided high resistance of the LSFs to channel errors, outperforming other schemes by, possibly, a considerable margin. Better objective and subjective performances were obtained with this new method which obviates the need for more powerful FEC schemes for transmission over noisy channels.<>     

A novel electrochemical process for the production of ammonium perchlorate

Perchloric acid, HCLO₄, and ammonium perchlorate, NH₄ClO₄, of high purity have been produced by electrolysis of chloric acid and subsequent reaction with high purity ammonium hydroxide to produce ammonium perchlorate. The process involves no alkali metals, chlorides or transition metals such as chromates, and thus produces propellant and explosive grade ammonium perchlorate of high purity and with no associated instability or pollution problems. The products can be recovered by solution crystallization-drying or direct spray-drying, respectively.     

Calculation of the acoustical characteristics of gas transport equipment of the unit type in the design stage

Translated from Khimicheskoe i Neftyanoe Mashinostroenie, No. 11, pp. 28–30, November, 1993.     

Impact and penetration by L/D ≤ 1 projectiles

Calculations of steel target penetration by L/D ≤ 1 tungsten and tungsten alloy projectiles have been extended to L/D = 1/32 over the velocity range 1.5 to 5 km/s. The ratio of crater to projectile diameter tends to 1 as L/D decreases over this entire velocity range. For impact velocities of 1.5 and 3 km/s, penetration depth normalized by projectile length, P/L, increases with decreasing projectile L/D up to a maximum value and then decreases for still lower L/D. Experiments at impact velocities of 2 and 3 km/s confirm these results. For 5 km/s impact velocity, the calculations show P/L increasing with decreasing projectile L/D over the entire range 1/32 ≤ L/D ≤ 1. The projectile L/D for which the maximum P/L occurs appears to depend on the impact velocity. P/L generally scales with impact velocity as P/L v^f(L/D) where f(L/D) ranges from 0 for a long rod to, we believe, 2 in the limit as projectile L/D approaches zero. The calculations show for 1/8 ≤ L/D ≤ 1/2, P/L v^0.9; for L/D = 1/16, P/L v^1.5; and for L/D = 1/32, the new results give P/L v^1.9.