Unisys Corporation

Prometrix Corporation

Anderson Jacobson, Inc.

Xetex, Inc.

Stanford Associates, Inc.

Radiation Detectors, Inc.

Philip D Wasserman

Chester L Mallory

Timothy R Corle

Terry R Seaver

A method for enhancing the performance of an artificially intelligent system employing a neural network by proving an optimized training set for training the neural network. The optimized training set is produced by identifying and removing inaccurate training pairs in the training set.

Method of enhancing the performance of a neural network

A method for enhancing the performance of an artificially intelligent system employing a neural network by proving an optimized training set for training the neural network. The optimized training set is produced by identifying and permanently removing inaccurate training pairs in the training set. The permanent removal of inaccurate training pairs is performed in a worst-error order.

Method of enhancing the selection of a training set for use in training

An improved real-time confocal scanning microscope, and an improved perforated disk for use in such microscope. A preferred embodiment of the inventive microscope includes a polarizing beamsplitting cube and a rotatable Nipkow disk perforted with a hexagonal hole pattern. The disk is preferably mounted so that the scan lines produced as the disk rotates will cross both the sample feature to be imaged and the sensor array in the system's video camera at an angle substantially equal to 45 degrees. This disk orientation ensures that brightness variations caused by a non-uniform scan will not affect the measurements. Rotation of the disk is preferably synchronized with the camera frame rate to prevent any scan errors from causing random (frame to frame) variations in the camera output. The polarizing beamsplitting cube consists of two triangular prisms connected (i.e., cemented) together by a dielectric film. Undesirable light reflections from the cube's faces are eliminated by orienting the cube about the dielectric film's normal axis so that light is incident at the cube at acute angles with respect to the cube's faces. An absorbing filter (such as a piece of black glass) is preferably optically cemented (or painted or otherwise coated) on the back face of the cube to absorb stray light.

Confocal scanning optical microscope

A nuclear radiation level detector includes an ionization chamber. Within the ionization chamber are spaced electrodes. The space between the electrodes communicates with the atmosphere so as to expose the electrodes to the atmosphere. A voltage source applies a potential across the electrodes. When the atmosphere is not contaminated with nuclear radiation above a reference level, the current flow between the electrodes is nil or substantially so. In the event the atmosphere is contaminated with nuclear radiation above a reference level, a measurable current flows between the electrodes. A measurement and alarm circuit operates an alarm in response to the current flow between the electrodes of a measurable magnitude.

Nuclear radiation level detector

A radiation level measuring system having a Geiger-Muller gas ionizing tube as a detector. The inherent dead-time of the Geiger-Muller tube that follows a discharge pulse is compensated by the pulse counting electronic measuring circuit. A signal pulse of varying duration is formed from the detector output pulses in a manner that corrects for this dead-time. Measurement of this generated signal pulse duration give the actual radiation level that would be measured if the detector dead-time did not exist.

Radiation flux measuring system with dead-time correction

A technique for reducing the nonlinear frequency characteristics of a carbon granule microphone of the type generally used in telephone handsets wherein a tone outside the voice range is applied to the microphone simultaneously with the audio signal desired to be transmitted. This technique has particular advantages with telephone modems, especially those where the sending and receiving frequencies are hamonically related.

Carbon microphone linearization technique

A function generator for producing digital data which represent a function of the time interval between two events and in which the digital data have a non-linear relationship to the time interval. The function generator comprises a rate multiplier circuit having a constant pulse frequency multiplicand and a continuously varying multiplier selected by a memory circuit throughout the time interval. A counter circuit has its input connected to the output of the rate multiplier circuit and has its output connected to the memory circuit to continuously address the memory circuit for selecting the continuously varying multiplier as each output pulse from the rate multiplier circuit is counted by the counter circuit. The data stored in the memory circuit for selecting the multiplier represents a series of straight line approximations to the desired continuous non-linear function. A digital feedback loop is formed between the rate multiplier circuit, the counter circuit and the memory circuit to cause a non-linear function to be generated by two events defining a time interval. At the completion of the time interval, the total count of pulses applied to the counter circuit is registered in a data storage device to provide digital data which represent a non-linear function of the time interval between two events.

Growing community of inventors

Cupertino, CA, United States of America

Philip D Wasserman