STATISTICAL ASSOCIATION LEARNING OF THE MARKOV DECISION PROCESS

A neural computing approach to the Markov decision problem is presented. The method makes use of historical records of rewards as inputs and average long-run incomes per epoch as targets for training a backpropagation network to associate the two quantities, without a priori knowledge of the state transition probabilities. Estimation of the long-run income, given a new reward matrix as input, is interpreted as a statistical-association learning problem. Following the training, the relevant functional relationship between the income (output) and the reward (input), that has been learned by the network, can be used to compute an unknown expected income. The present study examines network generalization to new inputs, considering the effects of different topological designs and the characteristics of training samples on neural computing accuracy. Successful performance of the best design, which is determined directly from the underlying mathematical model of the Markov decision problem, is demonstrated by a computer simulation experiment.     

Thermal layers near a friction surface of bodies in high-velocity slip

Thermal layers generated when one body slips over the surface of another are considered. Exactness of the solutions obtained by the methods of the boundary layer theory is analyzed.Institute of New Physical and Applied Problems of the Ukrainian Academy of Sciences, Kiev. Translated from Inzhenerno-fizicheskii Zhurnal, Vol. 63, No. 6, pp. 760–766, December, 1992.     

Self-aligning planarization and passivation for integration applications in III-V semiconductor devices

This work reports an easy planarization and passivation approach for the integration of III-V semiconductor devices. Vertically etched III-V semiconductor devices typically require sidewall passivation to suppress leakage currents and planarization of the passivation material for metal interconnection and device integration. It is, however, challenging to planarize all devices at once. This technique offers wafer-scale passivation and planarization that is automatically leveled to the device top in the 1-3-/spl mu/m vicinity surrounding each device. In this method, a dielectric hard mask is used to define the device area. An undercut structure is intentionally created below the hard mask, which is retained during the subsequent polymer spinning and anisotropic polymer etch back. The spin-on polymer that fills in the undercut seals the sidewalls for all the devices across the wafer. After the polymer etch back, the dielectric mask is removed leaving the polymer surrounding each device level with its device top to atomic scale flatness. This integration method is robust and is insensitive to spin-on polymer thickness, polymer etch nonuniformity, and device height difference. It prevents the polymer under the hard mask from etch-induced damage and creates a polymer-free device surface for metallization upon removal of the dielectric mask. We applied this integration technique in fabricating an InP-based photonic switch that consists of a mesa photodiode and a quantum-well waveguide modulator using benzocyclobutene (BCB) polymer. We demonstrated functional integrated photonic switches with high process yield of >90%, high breakdown voltage of >25 V, and low ohmic contact resistance of /spl sim/10 /spl Omega/. To the best of our knowledge, such an integration of a surface-normal photodiode and a lumped electroabsorption modulator with the use of BCB is the first to be implemented on a single substrate.     

Exciton photoluminescence in doped quasi-1D structures based on silicon

The dependence of carrier density in silicon quantum wires sheathed with SiO₂ on the wire diameter and the position of impurity atoms in respect to the wire center is analyzed theoretically. It is shown that, as the diameter of wires and nanocrystals decreases, the ionization energy of a dopant increases; therefore, the free carrier density decreases, and the screening of the Coulomb attraction becomes ineffective. As a result, the photoluminescence is defined by the radiative recombination of excitons even in the case of heavily doped Si. These conclusions are supported by the data of experimental study of spectral, excitation-power, and temperature dependences of photoluminescence in porous silicon structures fabricated on lightly and heavily doped Si substrates.     

The Influence of the Silicone Carbide Content and Type on the Microstructure and the Flaw Level of Silica Ceramics

Comprehensive studies of the flaw level and some structural characteristics of a composite ceramic material based on Karakeche clay (Kyrgyzstan) are carried out for a range of thermoactivation regimes and impurity contents of silicone carbides of various types.     

Statistical model of an undermoded reverberation chamber

Weibull distribution is adopted to model the electric field component of a Reverberation Chamber (RC). Its first property is to include the asymptotic laws, such as Rayleigh and exponential, and its main advantage lies in the fact that the Weibull shape parameter enables a model of the departure from overmoded to undermoded RC regime. Applications are given, such as an RC modal finite element modeling and a Monte Carlo simulation: they prove that the Weibull two-parameter distribution correctly models the quality factor influence. Moreover, the relevance of the use of this extreme value distribution is illustrated.     

Temperature Dependence of the Magnetic Properties of Amorphous Fe-Me-Si-B Alloys

Alloying of amorphous Fe-Me-Si-B metallic compositions with nickel and molybdenum results in a decrease in the values of saturation magnetization in fields with a magnetic intensity of 0.72 MA/m. The same alloying additions decrease the temperature of formation of new magnetic Fe₃B-type phases, which are stable up to 873 ± 5 K.     

Survey of nitrate and nitrite contents of vegetables grown in Korea.

A scientific basis for the evaluation of the risk to public health arising from excessive dietary intake of nitrate in Korea is provided. The nitrate () and nitrite () contents of various vegetables (Chinese cabbage, radish, lettuce, spinach, soybean sprouts, onion, pumpkin, green onion, cucumber, potato, carrot, garlic, green pepper, cabbage and Allium tuberosum Roth known as Crown daisy) are reported. Six hundred samples of 15 vegetables cultivated during different seasons were analysed for nitrate and nitrite by ion chromatography and ultraviolet spectrophotometry, respectively. No significant variance in nitrate levels was found for most vegetables cultivated during the summer and winter harvests. The mean nitrates level was higher in A. tuberosum Roth (5150 mg kg(-1)) and spinach (4259 mg kg(-1)), intermediate in radish (1878 mg kg(-1)) and Chinese cabbage (1740 mg kg(-1)), and lower in onion (23 mg kg(-1)), soybean sprouts (56 mg kg(-1)) and green pepper (76 mg kg(-1)) compared with those in other vegetables. The average nitrite contents in various vegetables were about 0.6 mg kg(-1), and the values were not significantly different among most vegetables. It was observed that nitrate contents in vegetables varied depending on the type of vegetables and were similar to those in vegetables grown in other countries. From the results of our studies and other information from foreign sources, it can be concluded that it is not necessary to establish limits of nitrates contents of vegetables cultivated in Korea due to the co-presence of beneficial elements such as ascorbic acid and alpha-tocopherol which are known to inhibit the formation of nitrosamine.