Showing papers by "Mitsubishi Electric published in 1992"

A Practical Approach to Feature Selection

Abstract: In real-world concept learning problems, the representation of data often uses many features, only a few of which may be related to the target concept. In this situation, feature selection is important both to speed up learning and to improve concept quality. A new feature selection algorithm Relief uses a statistical method and avoids heuristic search. Relief requires linear time in the number of given features and the number of training instances regardless of the target concept to be learned. Although the algorithm does not necessarily find the smallest subset of features, the size tends to be small because only statistically relevant features are selected. This paper focuses on empirical test results in two artificial domains; the LED Display domain and the Parity domain with and without noise. Comparison with other feature selection algorithms shows Relief's advantages in terms of learning time and the accuracy of the learned concept, suggesting Relief's practicality.

The feature selection problem: traditional methods and a new algorithm

Abstract: For real-world concept learning problems, feature selection is important to speed up learning and to improve concept quality We review and analyze past approaches to feature selection and note their strengths and weaknesses We then introduce and theoretically examine a new algorithm Rellef which selects relevant features using a statistical method Relief does not depend on heuristics, is accurate even if features interact, and is noise-tolerant It requires only linear time in the number of given features and the number of training instances, regardless of the target concept complexity The algorithm also has certain limitations such as nonoptimal feature set size Ways to overcome the limitations are suggested We also report the test results of comparison between Relief and other feature selection algorithms The empirical results support the theoretical analysis, suggesting a practical approach to feature selection for real-world problems

A new method for known plaintext attack of FEAL cipher

Abstract: We propose a new known plaintext attack of FEAL cipher. Our method differs from previous statistical ones in point of deriving the extended key in definite way. As a result, it is possible to break FEAL-4 with 5 known plaintexts and FEAL-6 with 100 known plaintexts respectively. Moreover, we show a method to break FEAL-8 with 215 known plaintexts faster than an exhaustive search.

Warning apparatus for a vehicle

Abstract: A warning apparatus for a vehicle generates a warning for a driver of the vehicle when the separation between the vehicle (1) and an obstacle (20) located in front of the vehicle falls below a prescribed value. The prescribed value is varied in accordance with changes in the physical or mental state of the driver of the vehicle, environmental conditions, or the driving characteristics of the driver and thereby adjust to changes in the driver's reaction time and the stopping distance of the vehicle.

Thermal Diffusivities of Composites with Various Types of Filler

Abstract: In-plane and out-of-plane thermal diffusivities (conductivities) of Kerimid resin composites reinforced with various types of filler were studied both experimentally and theoretically. The types of...

Fault diagnosis of power systems

Abstract: Knowledge-based fault diagnosis systems are classified by the method of inference utilized in the knowledge-based systems for fault diagnosis of power systems. The characteristics of each class and corresponding issues as well as the state-of-the-art techniques for improving their performance are presented. Additional topics covered are user interfaces, interfaces with energy management systems, and expert system development tools for fault diagnosis. Results and evaluation of actual operation in the field are also discussed. >

A 288-kb fully parallel content addressable memory using a stacked-capacitor cell structure

Abstract: A 288-kb (8 K words*36 b) fully parallel content addressable memory (CAM) LSI using a compact dynamic CAM cell with a stacked-capacitor structure and a novel hierarchical priority encoder is described. The stacked-capacitor structure results in a very compact dynamic CAM cell (66 mu m/sup 2/) which is operationally stable. The novel hierarchical priority encoder reduces the circuit area and power dissipation. In addition, a new priority decision circuit is introduced. The chip size is 10.3 mm*12.0 mm using a 0.8- mu m CMOS process technology. A typical search cycle time of 150 ns and a maximum power dissipation of 1.1 W have been obtained using circuit simulation. In fabricated CAM chips, the authors have verified the performance of a search operation at a 170-ns cycle and have achieved a typical read/write cycle time of 120 ns. >

Space voltage vector-based new PWM method for large capacity three-level GTO inverter

Abstract: A novel pulse-width-modulation (PWM) method for a three-level inverter based on space voltage vectors is presented. This PWM method can minimize the harmonic components of the output voltage under the minimum pulse-width limitation gate turn-off thyristors (GTOs) and can also suppress the fluctuation of the neural point DC voltage. Experimental results demonstrate that the proposed PWM method is suitable for a large-capacity three-level GTO inverter applied to motor drives for steel rolling mills, railcars, etc. >

Real-time digital simulator for power system analysis on a hypercube computer

Abstract: A digital simulator based on a hypercube-type massively parallel computer, the NCube2, has been developed. The simulator features: real-time simulation of a large power system which covers transient stability through long-term behavior with constant accuracy level in root mean square values; user-friendly man-machine interfaces which mimic the actual operating environment including interactive setting of several system parameters and real-time data presentation on a CRT; and high-speed A/D (analog to digital) converters, D/A (digital to analog) converters, and D/IO (digital input and digital output) interfaces are used to connect the digital simulator (power system dynamic model) to actual equipment. The authors describe the basic characteristics of the digital simulator and present results obtained in real-time simulations. >

A refreshable analog VLSI neural network chip with 400 neurons and 40 K synapses

Abstract: Describes a self-learning neural network chip with refresh on-chip analog synaptic weight storage. The chip integrates 400 neurons and 40000 synapses with 0.8- mu m double poly-Si double metal CMOS technology. Refresh time is less than 300 mu s. The chip retains learned information by repeating refresh at 100 ms intervals. The proposed refresh method is based on the decision made by a subnetwork. The subnetwork learns if the settling states of the main network should be memorized, retains the weights until they are relearned, and stores a 4-b representation of subnetwork weights in a counter. The main network is refreshed according to the output of the subnetwork. >

448-Detector optical recording system: development and application to Aplysia gill-withdrawal reflex

Abstract: The spatial resolution of a fast, multichannel, optical recording apparatus has been improved in an attempt to increase the completeness of optical recordings of neuron activity responsible for the Aplysia gill-withdrawal reflex. A new optical apparatus was developed, increasing the number of pixels to 448. Action potential activity from 168 to 192 neurons could be detected. Hence, the number of optically detected neurons increased about 2.5 times in comparison with a previously used 100- or 124-detector optical apparatus. Because optically detected action potential signals were shown to be due to the cell body of a neuron, it was possible to know its position as well as its approximate size. The distribution of the detected active neurons and those neurons that showed an apparent change in spike frequency to the siphon stimulation were determined. >

Source-to-drain nonuniformly doped channel (NUDC) MOSFET structures for high current drivability and threshold voltage controllability

Abstract: The source-to-drain nonuniformly doped channel (NUDC) MOSFET has been investigated to improve the aggravation of the V/sub th/ lowering characteristics and to prevent the degradation of the current drivability. The basic concept is to change the impurity ions to control the threshold voltage, which are doped uniformly along the channel in the conventional channel MOSFET, to a nonuniform profile of concentration. The MOSFET was fabricated by using the oblique rotating ion implantation technique. As a result, the V/sub th/ lowering at 0.4- mu m gate length of the NUDC MOSFET is drastically suppressed both in the linear region and in the saturation region as compared with that of the conventional channel MOSFET. Also, the maximum carrier mobility at 0.4- mu m gate length is improved by about 20.0%. Furthermore, the drain current is increased by about 20.0% at 0.4- mu m gate length. >

Controller for induction motor

Abstract: An error current component which becomes zero when an actual value of a primary flux coincides with a set value inducted by a product of an exciting current command and a primary self-inductance by detecting a primary current of an inductance motor, is inducted by an error current component processing circuit. A primary resistance compensation circuit processes a compensation amount of a primary resistance compensation amount. And a compensation voltage processing circuit processes a compensation voltage which directs the error current component to zero. A rotating speed of the induction motor can be controlled in a stable state even though a value of primary resistance of the induction motor is varied by temperature. Further, inadequate torque and excess current can be avoided.

Low threshold FS-BH laser on p-InP substrate grown by all-MOCVD

Abstract: A novel FS-BH laser structure using the selective growth characteristics of MOCVD on crystal facets has been developed. A threshold current as low as 12 mA and output power of 40 mW under CW operation have been achieved in the laser.

A 100-MHz 4-Mb cache DRAM with fast copy-back scheme

Abstract: A 4-Mb cache dynamic random access memory (CDRAM), which integrates 16-kb SRAM as a cache memory and 4-Mb DRAM into a monolithic circuit, is described. This CDRAM has a 100-MHz operating cache, newly proposed fast copy-back (FCB) scheme that realizes a three times faster miss access time over with the conventional copy-back method, and maximized mapping flexibility. The process technology is a quad-polysilicon double-metal 0.7- mu m CMOS process, which is the same as used in a conventional 4-Mb DRAM. The chip size of 82.9 mm/sup 2/ is only a 7% increase over the conventional 4-Mb DRAM. The simulated system performance indicated better performance than a conventional cache system with eight times the cache capacity. >

A new generation high speed low loss IGBT module

Abstract: By virtue of simulation and advanced process technologies the IGBT and the fast recovery diode chip structures have been modified to develop new generation low-loss power chips. These newly structured chips have been integrated into a moduler housing to form a 100A,600V dual type IGBT module that has exhibited superior characteristics, particularly in terms of reducing operational power loss over the conventional second generation counterpart. This paper will give details of this new generation IGBT module development and describe its characteristical superiority.

Measurement of SiH 2 Densities in an RF-Discharge Silane Plasmae Used in the Chemical Vapor Deposition of Hydrogenated Amorphous Silicon Film

Abstract: Silylene (SiH2) radicals have been detected by intracavity laser absorption spectroscopy in a parallel plate RF-discharge system used in the chemical vapor deposition of hydrogenated amorphous silicon. By the use of two different calibration methods, an absolute density ranging from 2×108 to 6×109 cm-3 has been obtained as a function of the input RF power and of the partial pressure of SiH4. The flux of SiH2 onto the substrate electrode estimated from the gradient of the spatial distribution is less than 1013 cm-2 s-1. This suggests that its direct contribution to the film deposition is small, although it plays an important role in gas phase reactions.

A mobility model including the screening effect in MOS inversion layer

Abstract: A mobility model for MOSFET device simulation is proposed. The model is not only applicable to both inversion layer and source/drain high concentration regions of a MOSFET, but it also takes into account the screening effect in the inversion layer. The model also includes an improved normal-field dependence for thin gate oxide MOSFETs. The low parallel electric field mobility is estimated by adding mobilities due to donor scattering, acceptor scattering and lattice scattering using Matthiesen's rule. Mobilities due to both the donor and the acceptor scattering include the electron screening effect. The mobility due to lattice scattering is formed as a function of normal electric field E/sub n/, including the strong dependence term of E/sub n/, to express surface roughness scattering. Calculation results of the device simulation using the mobility model show good agreement with the experimental data for various channel dopings. >

Shape Adjustment of a Flexible Space Antenna Reflector

Abstract: A recursive shape adjustment algorithm is proposed, which will be used for flexible antenna reflectors for space application. The antenna reflector is composed of flexible mesh surface, cable networks, and support structures. Length adjustment mechanisms are attached to the suspension wire between the surface cable and the backup network, and the shape of the mesh surface is controlled by the adjustment mechanisms. For the purpose of effective adjustment, the matrix of influence coefficients (sensitivity matrix) of the adjustment length to the surface displacements is needed, and the matrix is usually calculated by analytical model. But some modeling errors, such as prediction error of material properties or error due to model simplification, are unavoidable in the preliminary analysis. Besides, the sensitivity matrix may change during the process of adjustment because of nonlinearity of the structure in some cases. To avoid the above problems, a recursive shape adjustment algorithm is proposed. The major...

AlGaAs-GaAs pnpn differential optical switch

Abstract: A differential optical switch based on parallel-connected AlGaAs-GaAs pnpn structures is presented. Continuous input light makes it possible to discriminate between two optical inputs with a power difference of as low as 0.1 nW. The minimum operating energy with pulse inputs was 400 fJ. With the introduction of a storage operation of pnpn structures, the optical S-R flip-flop operation is also demonstrated with a 7.2-pJ optical input energy. The expansibility of the 2-D array is attractive in optical parallel processing systems, especially in optical neural networks. >

High‐speed template matching algorithm using information of contour points

Abstract: Template matching is a useful technique for matching images. This paper proposes a high-speed template matching technique which is effective for images that have sharp contours. The method uses only the image contour parts as one-dimensional templates. Although it can retain fine patterns, the data compression ratio and speed are very high. To improve reliability of matching, gray gradient information and intensity of contour points are used. Also, dilation of contours can reduce several errors due to disturbance of the contour or scaling. The algorithm herein can indicate differences between the template and the target image automatically after the alignment of images. The expanded algorithm applying the sequential similarity detection algorithm (SSDA) method also is discussed. Software simulation proved that the algorithm is 10 or 200 times faster than the conventional template matching technique which uses the cross-correlation estimation between two-dimensional (2-D) template and the target image. This algorithm was applied to special hardware. It can align two 512 × 512 pixels images in 40 ms when the size of the search area is 64 × 64.

Cyclic deformation of TiNi shape-memory alloy helical spring

Abstract: The influence of training and shape-memory processing temperature on cyclic deformation of TiNi shapememory alloy helical springs is investigated. The results are summarized as follows. (1) For large strain, significant irrecoverable elongation appears in the early cycles. The appearance of the irrecoverable elongation can be avoided by the training. (2) The recovery force of the spring varies slightly during the thermal cycles, which shows the very stable cyclic characteristics of the force. (3) The cyclic characteristics of deformation in the helical springs depend on the size of the region accompanied by the stress-induced martensitic transformation in the cross section of the wire.

Laser two-photon ionization detection of aromatic molecules on a metal surface in ambient air

Abstract: A laser two-photon process can ionize a photoabsorbing molecule selectively though a one-photon resonant process without ionizing the bulk surface. By measuring the ionized specles through photoconductivity, this technique has been found to be successful for highly sensitive detection of aromatic molecules on a metal surface in ambient air. The photocurrent was due to electron drift and oxygen anion drift in ambient air. The photocurrent signal of various aromatic molecules was approximately proportional to the absorbance at the laser wavelength

Guard rail detecting device

Abstract: A guard rail detecting device mounted on a motor vehicle comprises: two image sensors spaced (3, 4) vertically to sense a scene in front of the motor vehicle; window setting means (12) for setting a plurality of windows on the display screen (11) at predetermined positions; distance calculating means (10) for calculating the distances from the vehicle to the parts of the guard rail (5) which are held as images in the windows; means (10) for predicting the curvature of the guard rail from the distances thus calculated.

Mechanism for AlSiCu Alloy Corrosion

Abstract: We investigated corrosion at AlSiCu/TiN interconnection lines in VLSI devices. This type of corrosion appeared during the deionized water rinse treatment after the dry etching and ashing treatment (with O2+CF4 gas mixture). The corrosion shape was "pitlike". When added the vacuum heating treatment before the deionized water rinse, the corrosion number was decreased. However, the ashing treatment had a contrary effect for this type of corrosion suppression. We investigated this type of corrosion mechanism by ion-chromatography analysis. In this analysis, residual Cl concentration decreased when the ashing time was prolonged, but the corrosion number was increased significantly. We measured residual F concentration at the same time, and detected a high concentration of F- ions. We consider that these F- ions contribute to this type of corrosion.

Noncontacting card, terminal and transmission system therfore

Abstract: A noncontacting card 41 contains a coil circuit having at least two circuits of spiral coils 42, 47 connected to each other in parallel and electromagnetically coupled with a terminal 61, and a processing circuit 46 connected to the coil circuit for signal processing. This can perform high efficiency transmission of electric power and data from/to a terminal. A terminal 61 for a noncontacting card contains a coil circuit 62, 67 having at least two circuits of spiral coils connected to each other in parallel and electromagnetically coupled with a noncontacting card 41, and a processing circuit connected to the coil circuit for signal processing; this can perform high efficiency transmission of electric power and data to/from a noncontacting card. The card and terminal form a noncontacting transmission system. The spiral coils are of stranded wire. Selection means (71, 72) are provided for selecting a desired number of spiral coil wires and connecting these in parallel.

MGTP: A Parallel Theorem Prover Based on Lazy Model Generation

Abstract: We have implemented a model-generation based parallel theorem prover in KL1 on a parallel inference machine, PIM. We have developed several techniques to improve the efficiency of forward reasoning theorem provers based on lazy model generation. The tasks of the model-generation based prover are the generation and testing of atoms to be the elements of a model for the given theorem. The problem with this method is the explosion in the number of generated atoms and in the computational cost in time and space, incurred by the generation processes. Lazy model generation is a new method that avoids the generation of unnecessary atoms that are irrelevant to obtaining proofs, and to provide flexible control for the efficient use of resources in a parallel environment. With this method we have achieved a more than one-hundred-fold speedup on a PIM consisting of 128 PEs.