Evolving electronic circuits for computational intelligence hardware
This paper discusses the use of Evolvable Hardware (EHW) in automatic synthesis of electronic circuits for computational intelligence (CI) hardware.
Engineering topics
Publications and source records attributed to Jin, Y..
This paper discusses the use of Evolvable Hardware (EHW) in automatic synthesis of electronic circuits for computational intelligence (CI) hardware.
Explore the source record for details and available documents.
Explore the source record for details and available documents.
This paper describes the EHW development system, a tool that performs the evolutionary synthesis of electronic circuits, using the SPICE simulator and the Field Programmable Transistor Array hardware (FPTA) developed at JPL.
Evolvable Hardware (EHW) refers to HW design and self-reconfiguration using evolutionary/genetic mechanisms. The paper presents an overview of some key concepts of EHW, describing also a set of selected applications.
A translation sensitivity adjustable compact optical correlator (TSACOC) utilizing a convergent reference wave has been designed and tested.
Optical pattern recognition and neural associative memory are important research topics for optical computing.
In the Technology 2006 Case Studies/Success Stories presentation, we will describe and demonstrate a prototype of a compact optical pattern recognition system as an example of a successful technology transfer and continuuing development of state-of-the-art know-how by the close collaboration among government, academia, and small business via the NASA SBIR program. The prototype consists of a complete set of optical pattern recognition hardware with multi-channel storage and retrieval capability that is compactly configured inside a portable 1'X 2'X 3' aluminum case.
Solomon and van Tilborg have developed convolutional encoding algorithms for quadratic residue (QR) codes of lengths 47 and beyond. For these codes and reasonable constraint lengths, there are sequential decodings that may be simple, as in a convolutional encoding/decoding of the extended Golay Code. In addition, the previously found constraint length K = 9 for the (48, 24; 12) QR code was lower to K = 8 by Solomon. In our search for the smallest possible constraint lengths K for (80, 40; 16) self-dual quadratic residue and nonquadratic residue codes, we have found the constraint lengths K = 14 and K = 13, respectively. We have discovered a K = 21 convolutional encoding for the (104, 52; 20) QR code; there may be a smaller K for a (104, 52; 20) self-dual code that is not a quadratic residue code. The smaller the K, the less complex the sequential or Viterbi decoder.
A self-dual code length 48, dimension 24, with Hamming distance essentially equal to 12 is constructed here. There are only six code words of weight eight. All the other code words have weights that are multiples of four and have a minimum weight equal to 12. This code may be encoded systematically and arises from a strict binary representation of the (8,4;5) Reed-Solomon (RS) code over GF (64). The code may be considered as six interrelated (8,7;2) codes. The Mattson-Solomon representation of the cyclic decomposition of these codes and their parity sums are used to detect an odd number of errors in any of the six codes. These may then be used in a correction algorithm for hard or soft decision decoding. A (72,36;15) box code was constructed from a (63,35;8) cyclic code. The theoretical justification is presented herein. A second (72,36;15) code is constructed from an inner (63,27;16) Bose Chaudhuri Hocquenghem (BCH) code and expanded to length 72 using box code algorithms for extension. This code was simulated and verified to have a minimum distance of 15 with even weight words congruent to zero modulo four. The decoding for hard and soft decision is still more complex than the first code constructed above. Finally, an (8,4;5) RS code over GF (512) in the binary representation of the (72,36;15) box code gives rise to a (72,36;16*) code with nine words of weight eight, and all the rest have weights greater than or equal to 16.
Results are presented on the petrography and electron microprobe analyses of 14 discrete glass beads from the Luna 16 core sample (21036,15) from Mare Fecunditatis regolith, that were previously characterized as representing pristine glasses. Compared to Apollo pristine glasses analyzed by Delano (1986), the Luna 16 pristine glasses have higher CaO and Al2O3 contents but lower MgO and Ni. On the basis of their contents of MgO, FeO, Al2O3, and CaO, these pristine glasses could be divided into two groups, A and B. It is suggested that at least two parental magmas are needed to explain the chemical variations among these glasses. The Group B glasses appear to represent primitive parental magma that evolved by olivine fractionation to the compositions of the Luna 16 aluminous mare basalts, whereas the Group A volcanic glasses may represent an unusual new basalt magma type that contains a high plagioclase component.
Petrographic and electron microprobe techniques were used to examine igneous fragments in Luna 24 samples 24088 and 24105. It is the complex chemistry of the pyroxenes that distinguishes the different rock types. Basaltic pyroxenes exhibit an Fe-enrichment trend; the evolutionary trends are more complex in the gabbros, with enrichments in both Fe and Ti and a depletion in Cr. These chemical evolutionary trends are displayed by a progressive variation in rock types from Mg-rich olivine-gabbro to olivine-gabbro, and to ferrogabbro and ferrotroctolite. The low TiO2 content of the primary melt, possibly represented by the least-evolved Mg-rich olivine-gabbro, retarded the formation of early ilmenite and spinel, such that 'Fenner Trend' Fe enrichment occurred. The ferrotroctolite is probably the end product of chemical evolution by extreme fractional crystallization, controlled primarily by olivine and pyroxene crystallization.