R64-76 A Hybrid Analog-Digital Parameter Optimizer for Astrac II
760
IEEE TRANSACTIONS ON ELECTRONIC
COMPUTERS
December
tion problem, and in the combinatorial minimization procedure used. The authors used the procedure described in an article by Prather2 in which Urbano and Mueller's method of finding prime implicants3 is programmed. This method is not particularly better or worse for the present application than alternative procedures. Unimportant details are included such as the role played by three DO loops in generating the input line equations, and many ungrammatical sentences are to be found in the text. In summary, this article will be incomprehensible to the reader who has not read Phister's book, and it will not be of interest to the reader who has read chapter five of this book. S. R. PETRICK AF Cambridge Research Labs. Bedford, Mass.
J function," . Assoc. for Comp. Mach., vol. 7; October, 1960. R. H. Urbano and R. K. Mueller,"A topological method for the determination of the minimal forms of a Boolean function," IRE TRANS. ON ELECTRONIC COMPUTERS, 126-132; September, 1956. Vol. EC-5, 3
2
"Computational R. Prather,
aids for determining the minimal form of
truth
pp.
E. ANALOG AND HYBRID SYSTEMS R64-76 A Hybrid Analog-Digital Parameter Optimizer for Astrac IIBaker A. Mitchell, Jr. Proc. A FIPS 1964 Spring Joint Computer Conf., vol. 25. This paper demonstrates very well how high-speed repetitive analog computers can be combined with digital logic and hybrid ele-
Optimization is obtained by alternately varying the set of parameters and solving for the index of performance. Parameters are set by counter controlled DIA converters used as multipliers and remain set during the compute cycle. They are retained if an improvement is realized, otherwise not. Although the equipment discussed is capable of being used with deterministic methods, the paper is cerned with the implementation of a modified random search. con-
ments to form highly practical systems.
fixed vectors, and the succession of segments is represented by a sequence of octal digits. Since mating between two pieces usually occurs over some fraction of their respective boundaries, the boundary of each piece is partitioned into sections (called"chainlets"), each of which is likely to mate with one chainlet from another piece. The guide used to form these chainlets is the occurrence of slope discontinuities as mentioned above. Solution of the jigsaw puzzle is now a matter of mating chainlets. Rather than making an exhaustive search, the program first classifies the chainlets according to some of their measurable, orientationinvariant"features" (e.g., length of the chainlet, length of the straight line between the end points, etc.). Chainlets having similar features are grouped as possible mating candidates, thus narrowing the required search. While puzzle assembly can take place in many ways, the authors chose a method based on outward spiralling growth centered around one piece. Assembly errors, which may arise as the result of quan
tization noise (introduced in forming the straight-line approximation to the piece contours) or ambiguities (a mating which appears valid but is subsequently shown to be in error) are kept under control by keeping track of alternate mating possibilities. When an inconsistency is detected, the most likely alternate possibilities can be tried. In the paper described above, the authors have defined an interesting problem and have obtained a"workable," if not elegant, solution. Although it is claimed that this study was chosen as a vehicle to develop computer techniques for application to pattern recognition problems and game solving, it is difficult to see where this paper makes a contribution outside of the context of the given problem. MARTIN A. FISCHLER Lockheed Missiles& Space Co. Palo Alto, Calif.
The outputs from two independent noise generators are connected 4 -, and 3 probabilities to bias to combinatorial circuits to achieve, the random walk. Two memory elements per parameter used to hold logical decisions until all parameters are set and can also be used along with a success/failure indication to further bias the walk. The system described was built and the results of several searches are shown. No relative economic comparisons are made. This paper, although it does not contribute to the theory of optimization, does show how relatively uncomplicated circuits can be combined to produce a system where new and rather exciting techniques may be uncovered. The reviewer recognizes, however, no panacea to the search problem in spaces of high dimensions or with multi-modal functions. DON J. NELSON
G. REDUNDANCY TECHNIQUES R64-78 Interwoven Redundant Logic-W. H. Pierce. Inst., vol. 277, pp. 55-85; January, 1964.
J.
Franklin
In this paper, W. H. Pierce introduces two types of redundancy schemes for error correction in gate-type logical nets: 1) an alternating-layer-error-correcting scheme; and 2) a single-layer-errorcorrecting scheme. These schemes may be generally described as follows: For each gate F =
Director of the Computing Center
University of Nebraska Lincoln, Neb.
F. PATTERN RECOGNITION R64-77 Apictorial Jigsaw Puzzles: The Computer Solution of a Problem in Pattem Recognition-H. Freeman and L. Garder. IEEE TRANS. ON ELECTRONIC COMPUTERS, vol. EC-13, pp. 118-127; April, 1964. This paper describes a procedure that enables a digital computer to solve"apictorial" jigsaw puzzles, i.e., puzzles in which all pieces are uniformly gray and the only available information is the shape of the pieces. The main problems discussed here are: 1) the description of the pieces (coding and classification), and 2) puzzle assembly and evaluation of correct fit. The authors describe a wide variety of possible puzzle types, including puzzles with irregular exterior boundaries and with extra, or missing, pieces. However, they are mainly concerned with what happens at the points (called"junction points") where the boundary between two puzzle pieces dive
rges into two or more separate curves.
computing f(xi, x,,,,)=y, construct a x1i;; redundancy-modified gate F that computesf(xl, x12, x_==XiB=xi, I, xmB) f(Xl,* * *, xm) for x,'= a xm1, xm2, i 1, 2,* *, m. For each i, 1<i< m, let the set of redundant signals, xii,. xiB be called a bunch, denoted by Xi. For example, in the case of a NAND gate computing xl X2=y, the corresponding redundancy-modified-NAND gate (RM-NAND gate) will compute Xll X12- X21- X22 The following error characteristics of the RM-NAND gate are of interest: 1) The RM-NAND gate is insensitive to a single (0-1) error in any one of its lines. (The value ofy will not change as a result of such an error.) Pierce refers to such an error as a subcritical error (S-error). Two subcritical errors occurring in the same bunch of a RM-NAND gate will cause a (1-*0) output error for some input valuation. However, if the two S-errors are distributed one per bunch, the output will not change its value as a result of the errors. 2) A (1 -+0) error inoneof itsinputlineswill cause a (0-4) output error for the input valuation: xi=1 and X2 1. Such an error is called by Pierce a critical error (C-error). Now, if one RM-NAND gate feeds another RM-NAND gate, then a C-error at the input of the first will only appear as an S-error at the input of the second. The C-error may, therefore, be corrected by the second RM-NAND gate, provided that the second gate does not have two S-errors in the , =
=
same
bunch.
The approach presented here is mainly designed to solve those puzzles in which junction points are relatively easy to distinguish from other points on the boundaries of the puzzle pieces (e.g., in many puzzles, slope discontinuities almost always occur at such a junction point and are infrequent elsewhere). A"chain-encoding" technique is used to encode the piece descriptions. This scheme produces a straight-line-segment approximation to the given boundary. The segments are chosen from a set of eight
correcting" if the redundant signals are so distributed that the collection of C-errors to be corrected, occurring in one layer of the net, does not cause simultaneous S-errors in all the lines of any one bunch feeding the next layer of RM-NAND gates. Similar considerations also apply for the RM-NOR gate nets. In the of (AND, OR) gate nets, alternation of C- and S-errors is obtained by alternating RM-AND and RM-OR gates, each RMAND gate feeding an RM-OR gate and vice versa. case
An RM-NAND gate net will thus be"alternating-layer-error-

63R64中文资料 隐藏>> 元器件交易网www.cecb2b.com Optical Encoders SERIES 63R High Resolution, Ball Bearing, 5-pin (Polarized Connection) FEATURES ? 25...
基于HT46R64的数字定时器设计基于HT46R64的数字定时器设计隐藏>> 90 基于HT46R64的数字定时器设计 基于HT46R64的数字定时器设计术Design 吴洪江on HT46R64Based...
关键词:46R64 1/2 相关文档推荐 典型电路 21页 1财富值 典型电路 7页 2财富值 典型电路 11页 免费 典型电路1 84页 5财富值 典型题解 电路 83页 5财富...
M3R64FCJ中文资料_电子/电路_工程科技_专业资料。元器件交易网www.cecb2b.com M3R Series 9x14 mm, 3.3 Volt, HCMOS/TTL, Clock Oscillator Pin Connections...
HT46R64中文资料_电子/电路_工程科技_专业资料 暂无评价0人阅读0次下载举报文档 HT46R64中文资料_电子/电路_工程科技_专业资料。移动电源设计资料...
ht46r64b头文件_电子/电路_工程科技_专业资料。合泰单片机// //Holtek C header file. //File name: HT46R064B.H // #ifndef _HOLTEK_H_ #define _...
为什么以太网的最小数据帧为64字节_计算机硬件及网络_IT/计算机_专业资料。数据帧为什么以太网数据帧最小为 64 字节如果把“以太网”比作是一栋“房子”,这栋...
(R) ,此时弹出如下界面: PDF 文件使用 "pdfFactory Pro" 试用版本创建 www.fineprint.cn 点击 浏览(R)找到驱动放置的根目录,选择 usbasp 驱动(x64)64 位 ...
64 位 WIN7 系统搭建 Android 开发环境 1、在原来的电脑上直接运行 eclpise,但是虚拟机总是无法打开,所以把以前 eclipse 删 掉了,在官网上下载最新的 eclipse,...
热门文档