Abstract: A sensor system includes a transmitter configured to emit electromagnetic radiation towards a portion of an environment and an optical neural network configured to receive a reflection of the electromagnetic radiation from the portion of the environment and generate an array of electromagnetic signals. A property of each respective electromagnetic signal of the array of electromagnetic signals represents a corresponding physical characteristic of the portion of the environment. The sensor system also includes a detector array configured to receive the array of electromagnetic signals and including a plurality of electromagnetic signal detectors. Each respective electromagnetic signal detector is configured to measure the property of a corresponding electromagnetic signal of the array of electromagnetic signals and generate, based on the measured at least one property of the corresponding electromagnetic signal, a value representing the corresponding physical characteristic.

Abstract: An accelerator for modern convolutional neural networks applies the Winograd filtering algorithm in a wavelength division multiplexing integrated photonics circuit modulated by a memristor-based analog memory unit.

Abstract: Most artificial neural networks are implemented electronically using graphical processing units to compute products of input signals and predetermined weights. The number of weights scales as the square of the number of neurons in the neural network, causing the power and bandwidth associated with retrieving and distributing the weights in an electronic architecture to scale poorly. Switching from an electronic architecture to an optical architecture for storing and distributing weights alleviates the communications bottleneck and reduces the power per transaction for much better scaling. The weights can be distributed at terabits per second at a power cost of picojoules per bit (versus gigabits per second and femtojoules per bit for electronic architectures). The bandwidth and power advantages are even better when distributing the same weights to many optical neural networks running simultaneously.

Abstract: Systems and methods that include: providing input information in an electronic format; converting at least a part of the electronic input information into an optical input vector; optically transforming the optical input vector into an optical output vector based on an optical matrix multiplication; converting the optical output vector into an electronic format; and electronically applying a non-linear transformation to the electronically converted optical output vector to provide output information in an electronic format. In some examples, a set of multiple input values are encoded on respective optical signals carried by optical waveguides. For each of at least two subsets of one or more optical signals, a corresponding set of one or more copying modules splits the subset of one or more optical signals into two or more copies of the optical signals.

Abstract: The present invention relates to a system and method for recognizing a consonant and improving recognition performance within a short time, compared to prior art requiring a frequency decomposition process, by performing frequency analysis and voice recognition in real time without undergoing a frequency decomposition process of voice.

Abstract: An electro-optic beam controller, material processing apparatus, or optical amplifier, and corresponding methods, can include an actively controlled, waveguide-based, optical spatial mode conversion device. The conversion device can include a coupler, which can be a photonic lantern, configured to combine light beams into a common light beam; a sensor configured to measure at least one characteristic of the common light beam; and a controller configured to modulate optical parameters of the individual, respective light beams to set one or more spatial modes of the common light beam. Actively controlled and modulated devices can be used to maintain a stable, diffraction-limited beam for use in an amplification, communications, imaging, laser radar, switching, or laser material processing system. Embodiments can also be used to maintain a fundamental or other spatial mode in an optical fiber even while scaling to kilowatt power.

Abstract: This disclosure includes methods for designing a simplified Integrated Computational Element (ICE) and for optimizing a selection of a combination of ICE designs. A method for fabricating a simplified ICE having one or more film layers includes predicting an optimal thickness of each of the one or more film layers of the simplified ICE using a neural network. A method for recalibrating the fabricated ICE elements for system implementation is also disclosed. The disclosure also includes the simplified ICE designed by and the ICE combination selected by the disclosed methods. The disclosure also includes an information handling system with machine-readable instructions to perform the methods disclosed herein.

Abstract: Disclosed embodiments relate to a variable format, variable sparsity matrix multiplication (VFVSMM) instruction. In one example, a processor includes fetch and decode circuitry to fetch and decode a VFVSMM instruction specifying locations of A, B, and C matrices having (M×K), (K×N), and (M×N) elements, respectively, execution circuitry, responsive to the decoded VFVSMM instruction, to: route each row of the specified A matrix, staggering subsequent rows, into corresponding rows of a (M×N) processing array, and route each column of the specified B matrix, staggering subsequent columns, into corresponding columns of the processing array, wherein each of the processing units is to generate K products of A-matrix elements and matching B-matrix elements having a same row address as a column address of the A-matrix element, and to accumulate each generated product with a corresponding C-matrix element.

Abstract: A system and method are disclosed for managing learning facilities in a modular learning system. Information about a plurality of learning facilities, each associated with learning infrastructure, is stored in the modular learning system. Learning applications are also stored in the modular learning system, with each learning application associated with ergonomic requirements. The modular learning system receives a request to set up a learning facility for at least one learning application, and based on the infrastructure and ergonomic requirements determines whether the facility and the at least one learning application are compatible.

Abstract: A method for manufacturing a polymer dispersed liquid crystal (PDLC) panel, which comprises: obtaining a liquid crystal mixture by uniformly mixing liquid crystal molecules, a photoinitiator and a prepolymer; forming a liquid crystal cell (23) by injecting the obtained liquid crystal mixture into an empty liquid crystal cell (22); and obtaining the PDLC panel by adopting uniformly alternated intense and weak light to irradiate the liquid crystal cell (23) to allow the liquid crystal mixture in the liquid crystal cell (23) to be subjected to polymerization. By adoption of the method, the manufactured PDLC panel has uniform brightness.

Abstract: A cellular computational platform is disclosed that includes a multiplicity of functionally identical, repeating computational hardware units that are interconnected electrically and optically. Each computational hardware unit includes a reprogrammable local memory and has interconnections to other such units that have reconfigurable weights. Each computational hardware unit is configured to transmit signals into the network for broadcast in a protocol-less manner to other such units in the network, and to respond to protocol-less broadcast messages that it receives from the network. Each computational hardware unit is further configured to reprogram the local memory in response to incoming electrical and/or optical signals.

Abstract: A plurality of chips arranged in a certain layout so as to face free space, and one or more optical elements are included. In the case where signal traffic for electrical communication with a given chip exceeds or is expected to exceed a certain threshold, a plurality of chips involved in communication routing of the excess signal traffic are identified, part of related signal traffic that crosses the plurality of identified chips is converted from an electric signal into an optical signal to re-route the excess signal traffic, and paths of the related signal traffic are dynamically adapted from fixed wired paths between the plurality of chips to optical communication paths formed in the free space.

Abstract: A plurality of chips arranged in a certain layout so as to face free space, and one or more optical elements are included. In the case where signal traffic for electrical communication with a given chip exceeds or is expected to exceed a certain threshold, a plurality of chips involved in communication routing of the excess signal traffic are identified, part of related signal traffic that crosses the plurality of identified chips is converted from an electric signal into an optical signal to re-route the excess signal traffic, and paths of the related signal traffic are dynamically adapted from fixed wired paths between the plurality of chips to optical communication paths formed in the free space.

Abstract: Neuronal networks of electronic neurons interconnected via electronic synapses with synaptic weight normalization. The synaptic weights are based on learning rules for the neuronal network, such that a synaptic weight for a synapse determines the effect of a spiking source neuron on a target neuron connected via the synapse. Each synaptic weight is maintained within a predetermined range by performing synaptic weight normalization for neural network stability.

Abstract: A method for assessing the integrity of a rock mass, the method including impacting the rock mass, capturing an acoustic signal generated as a result of the impact, deriving a frequency distribution for the captured acoustic signal, processing data from the frequency distribution by means of a neural network process applying artificial intelligence to assess the inputted data, and presenting a signal from the neural network process which is indicative of the integrity of the rock mass.

Abstract: A system and method for optimally determining appropriate ergonomics for an individual in a workspace, such as an office, a vehicle, or a school, having at least one item with at least one physically adjustable parameter comprises collecting input data; determining at least one preferred setting for the at least one physically adjustable parameter based at least in part upon the input data collected in the survey; and adjusting the at least one physically adjustable parameter of the at least one item to the preferred setting based upon the results of the determination. The input data can comprise information related to the individual, a task performed by the individual, an environmental feature of the workplace, and/or to a comfort level of the individual. Further, the input data can be collected through the Internet.

Abstract: In a procedure for regulating a combustion process in an installation while air is being supplied, material is converted by the combustion process, with at least one flame being formed, and the state variables (s(t)) describing the state of the system in the installation are determined using at least one observation device that images the flame, as well as other sensors, and are evaluated in a computer, whereupon any appropriate actions (ai) that may be needed are selected in order to control adjusting devices for the supply of material and/or air, wherein during setpoint regulation to achieve setpoints (s0) of the state variables and/or stability of the combustion process a changeover is occasionally made from setpoint control to disturbance control.

Abstract: We develop a system consisting of a neural architecture resulting in classifying regions corresponding to users' keystroke patterns. We extend the adaptation properties to classification phase resulting in learning of changes over time. Classification results on login attempts of 43 users (216 valid, 657 impersonation samples) show considerable improvements over existing methods.

Abstract: An indicia for marking on an object, such as a cartridge case, for representing selected information, such as identification indicia identifying the firearm that discharged the cartridge, and methods and apparatus for generating, imprinting and reading the identification indicia. An indicia includes a multi-dimensional array of encoded marks, including encoded marks determined by spectral encoding variables representing the selected information wherein each spectral variable is spectrally distinguishable from others of the spectral variables representing variables, and an encoded pattern of the encoded marks is determined by algorithmic transformation of the selected information. An indicia may also include hologram related artwork. An indicia may be an encoded hologram multi-dimensional barcode, an encoded hologram or an encoded concentric circular barcode and may be formed of a single indicia or as an array of indicia.

Abstract: A method and system for facilitating resolution of engineering and business issues is provided which comprises an issue component for identifying the issues to be resolved, an inquiry component for facilitating collection of client information relevant to the issue component to facilitate definition of the issue component, a knowledge base comprising data and information for facilitating assessment of the client information; and a solution base for compiling assessments and recommendations from the knowledge base and for reporting the assessments and recommendations to a client. In addition, the system can include a network for communicating the client information to the knowledge base and for communicating the assessments and recommendations to the client.

Abstract: A novel microelectromechanical system artificial neural network (MEMS ANN) device performs the function of a conventional artificial neural network node element. Micro-machined polysilicon or high aspect ratio composite beam micro-resonators replace as computational elements the silicon transistors and software simulations of prior-art ANNs. The basic MEMSANN device forms a non-linear (e.g., sigmoid) function of a sum of products. Products of the magnitudes of sine waves, applied to the input drive comb and shuttle magnitudes, are formed in the frequency domain and summed by coupling a plurality of resonators with a mechanical coupling frame, or by integrating them into one resonator. A sigmoid function is applied to the sum of products by shaping the overlap capacitance of the output comb fingers of the resonator. Methods of building and using various single MEMS ANN devices and multi-layered arrays of MEMS ANN circuits are also described.

Abstract: The present invention discloses low noise, optically coupled optoelectronic and all-optical artificial neuron devices that can be configured in an array to simulate the function of biological neural networks, and methods for making the artificial neurons. In a first optoelectronic embodiment, the device employs the regenerative pulsation property of astable multivibrators as optical pulse generators. A photosensitive detector, disposed in a circuit to control the state of an astable or bistable multivibrator, converts the intensity of the input light into a train of light pulses having a frequency that is a function of the intensity of the input signal. In an all-optical embodiment of an artificial neuron, an input signal is first integrated and the integrated signal transmitted to an optical pulse generator comprised of a nonlinear material disposed within the cavity of a Fabry-Perot etalon.

Abstract: The present invention discloses low noise, optically coupled optoelectronic and all-optical artificial neuron devices that can be configured in an array to simulate the function of biological neural networks, and methods for making the artificial neurons. In a first optoelectronic embodiment, the device employs the regenerative pulsation property of astable multivibrators as optical pulse generators. Prior art pulse-coupled artificial neurons are subject to undesirable noise interference because the interconnection of such prior art neurons is based on electrical signals conducted through a grid of wires. The present invention obviates the need for hard-wired interconnection of individual neurons in order to cofigure the neurons into a network. In an optoelectronic embodiment, the neuron receives an optical input signal from an external source.

Abstract: A Hebbian synapse emulation circuit models the conductance of a synapse circuit. In one embodiment, the circuit includes a counter that provides control signals determining the conduction states of electrical pathways, which define the conductance level of the synapse. The counter can increment, decrement, or leave the same, the synapse conductance value based upon potentiation and depression signals that are derived from a voltage that corresponds to the calcium concentration in the synapse. The counter can be coupled to a plurality of synapse circuits on a time-shared basis.

Abstract: In optical filter systems and optical transmission systems, an optical filter compresses data into and/or derives data from a light signal. The filter way weight an incident light signal by wavelength over a predetermined wavelength range according to a predetermined function so that the filter performs the dot product of the light signal and the function.

Abstract: A neural network circuit is provided having a plurality of circuits capable of charge storage. Also provided is a plurality of circuits each coupled to at least one of the plurality of charge storage circuits and constructed to generate an output in accordance with a neuron transfer function. Each of a plurality of circuits is coupled to one of the plurality of neuron transfer function circuits and constructed to generate a derivative of the output. A weight update circuit updates the charge storage circuits based upon output from the plurality of transfer function circuits and output from the plurality of derivative circuits. In preferred embodiments, separate training and validation networks share the same set of charge storage circuits and may operate concurrently.

Abstract: A glass break detector is disclosed that uses a neural network to determine if an audio signal is breaking glass. A characteristic extraction unit is used to extract a set of signal characteristics from a time domain signal based on the audio signal. The set of signal characteristics is the set of the magnitudes of the discrete Fourier transform coefficients of an acquired time domain signal, or the Fourier transform coefficients themselves. A classifier is connected to the characteristic extraction unit. It is a two-layer neural network that uses the set of signal characteristics to accurately determine whether the acquired time domain signal represents breaking glass.

Abstract: The invention is made of an apparatus which incorporates a sensor-based means for stabilizing random access networks. In the preferred embodiment, a grid of sensors is used to gather energy measurements for analysis. In a preferred embodiment a neural network has been trained to estimate the number of colliding users in a given slot. This information is used to set parameters in a backoff algorithm so as to stabilize the network and minimize the delay experienced by users. The invention has the ability to locate users geographically within the network coverage area. This information can be used in conjunction with a steerable beam or an array of antennas to develop geographically-determined Aloha subchannels, further increasing the capacity of the system.

Abstract: The e-cell is a computational element for the construction of circuits capable of performing transform invariant pattern recognition, scene segmentation, and regularity extraction in a variety of sensory modes including vision, hearing, olefaction, touch, etc. The e-cell consists of a cell body, an output terminal, and set of tree-connected input regions each of which have one or more input terminals and corresponding stored weights. The e-cell is capable of learning and subsequently discriminating multidimensional (in the mathematical sense) activation patterns applied to its inputs. The strictness with which the e-cell requires the presence in the target of each dimension of a learned pattern is dynamically adjustable, thus making it suitable to recognition of learned patterns under occlusion and various kinds of degradation. The strictness of discrimination across the region dimension is independent of the strictness of recognition within each dimension.

Abstract: An algorithm method of predicting estimated sea energy, wave directions and other seastate data with respect to submerged sea-going vessels, based on inputs derived from signal processed measurements of keel depth, pitch, roll and forward speed applied to three neural networks for heading detection and to a fourth neural network for seastate estimations.

Abstract: A self-routing switch such as a Banyan switch has a controller which recognizes incoming routing requests which would give rise to blocking in the switch and makes and optimum selection of queued requests which can be handled without blocking. The controller is implemented by means of an optical neural network having a light source array to illuminate a photodetector array through a mask, there being a light source array element and a photodetector element for each possible path through the switch. The assignment of paths to array locations is such that, for the path corresponding to any light source array element, the photodetector array element positions corresponding to the paths blocked thereby form a pattern which is a shifted version of the pattern formed by the photodetector array elements corresponding to the paths blocked by a path corresponding to any other light source array element, whereby a single mask may be employed.

Abstract: A solder joint inspection process and system includes using a correlation values to classify the shape of the solder joint. The solder joint is illuminated by a multiple colors, and the image of the solder joint is captured. This multiple color image is then converted into a plurality of one-dimensional vectors. Each one-dimensional vectors sequence is for one color of the multi-color illumination and for a one-dimensional spatial extent. Correlation values among all combinations of the one-dimensional vector sequences are computed and used for training an automatic solder joint shape classifier.

Abstract: A method and apparatus for an electronic artificial neural network, which serves as an associative memory that has a complete set of N-dimensional Hadamard vectors as stored states, suitable for large N that are powers of 2. The neural net has nonlinear synapses, each of which processes signals from two neurons. These synapses can be implemented by simple passive circuits comprised of eight resistors and four diodes. The connections in the neural net are specified through a subset of a group that is defined over the integers from 1 to N. The subset is chosen such that the connections can be implemented in VLSI or wafer scale integration. An extension of the Hadamard memory causes the memory to provide new Hadamard vectors when these are needed for the purpose of Hebb learning.

Abstract: Each neural element of a column-structured recurrent neural network generates an output from input data and recurrent data provided from a context layer of a corresponding column. One or more candidates for an estimated value is obtained, and an occurrence probability is computed using an internal state by solving an estimation equation determined by the internal state output from the neural network. A candidate having the highest occurrence probability is an estimated value for unknown data. Thus, the internal state of the recurrent neural network is explicitly associated with the estimated value for data, and a data change can be efficiently estimated.

Abstract: A vector normalizing apparatus which normalizes an input vector or weight vectors by L.sub.2 -norm and adds norm information to the vector before it is normalized. Vector data from an input vector acquiring device (1) is displayed on an input vector display device (2) and transmitted through a vector transmitting device (3) to a normalized vector output device (4). During the transmission of the vector data, L.sub.2 -norm of the vector data is calculated by an L.sub.2 -norm calculating device (5) including a vector component squaring device (7) for squaring each of the components of the displayed vector, a summation device (8) for calculating a total sum of the squared components, and a square-root calculating device (9) for calculating L.sub.2 -norm by obtaining a square root of the result of the summation. The input vector is normalized by a vector component adjusting device (6) using the calculated value of L.sub.2 -norm. Thus, the input vector normalized by the L.sub.

Abstract: An optoelectronic cellular neural network, which can be programmed optically, in amorphous or polycrystalline silicon, which makes up a monolithic image processing system with optical input and output which can be programmed by optical signals. Its layered structure for both the upper and lower surfaces are occupied, without discontinuity, by photosensitive and photoemissive devices. The network basically consists of an input photosensitive layer, a processing layer, a photosensitive control layer and also an optical control mask.

Abstract: A compact neural network architecture is trainable to sense and classify an optical image directly projected onto it. The system is based upon the combination of a two-dimensional amorphous silicon photoconductor array and a liquid-crystal spatial light modulator. Appropriate filtering of the incident optical image upon capture is incorporated into the net work training rules, through a modification of the standard backpropagation training algorithm. Training of the network on two image classification problems is described: the recognition of handprinted digits, and facial recognition. The network, once trained is capable of standalone operation, sensing an incident image and outputting a final classification signal in real time.

Abstract: An image recognition device includes an input section for inputting an image; a pattern extraction section for extracting a plurality of predetermined patterns from the input image; a first feature extraction section for extracting a first feature indicating the direction in which the first pattern of the extracted predetermined patterns exists with respect to a point on a second pattern of the extracted predetermined patterns; a second feature extraction section for extracting a second feature indicating an amount of the first pattern existing in the direction with respect to the second pattern, based on the first feature; and a recognition section for recognizing the input image based on the second feature.