Qualcomm Incorporated

Intel Corporation

Somdeb Majumdar

David Jonathan Julian

Gene Wesley Marsh

Chong U Lee

Edward Harrison Teague

Harinath Garudadri

Pawan Kumar Baheti

Regan Blythe Towal

Chinnappa K Ganapathy

William Henry Von Novak, Iii

Venkata Sreekanta Reddy Annapureddy

Dexu Lin

Muhammed Ibrahim Sezan

Daniel Keyes Butterfield

Sanjay Marthanda

Jeffrey Alexander Levin

Ernest Tadashi Ozaki

Robert S Tartz

Mattias Marder

Venkat Rangan

Mainak Biswas

Anthony Sarah

Rinat Burdo

Yi Tang

Aaron Douglass Lamb

Adnan M Agbaria

Aniket Arun Vartak

Lior Storfer

Marcel Nassar

Sachin Subhash Talathi

Adrian J Prentice

Karamvir Chatha

Xin Wang

Robert William Marshall

Mohammed Shoaib

Estelle Aflalo

William Richard Bell, Ii

M Anthony Lewis

Amin Fazeldehkordi

Tristan Webb

Richard Anthony Calle

Edward Liao

Ron Banner

Rouzbeh Kashef

Hanlin Tang

Santiago Miret

Avrech Ben-David

David Edward Howard

Shauharda Khadka

Evren Tumer

Ii William Richard Bell

Pawan K Baheti

Automatic multi-objective hardware optimization for processing a deep learning network is disclosed. An example of a storage medium includes instructions for obtaining client preferences for a plurality of performance indicators for processing of a deep learning workload; generating a workload representation for the deep learning workload; providing the workload representation to machine learning processing to generate a workload executable, the workload executable including hardware mapping based on the client preferences; and applying the workload executable in processing of the deep learning workload.

Automatic multi-objective hardware optimization for processing of deep learning networks

An apparatus for classifying an input includes a classifier and a feature extractor. The feature extractor is configured to generate a feature vector based on the input. The feature vector is also configured to set a number of elements of the feature vector to zero to produce a sparse feature vector. The sparse feature vector has the same dimensions as the feature vector generated by the feature extractor. However, the sparse feature vector includes fewer non-zero elements than the feature vector generated by the feature extractor. The feature vector is further configured to forward the sparse feature vector to the classifier to classify the input.

Generating a sparse feature vector for classification

Logic may quantize a primary neural network. Logic may generate, by a secondary neural network logic circuitry for a primary neural network logic circuitry, quantization parameters. The primary neural network logic circuitry may comprise a primary neural network with multiple layers trainable with an objective function. Each of the multiple layers of the primary neural network may comprise multiple tensors. The secondary neural network logic circuitry may comprise one or more secondary neural networks trainable with the objective function to output the quantization parameters to the tensors.

Methods and arrangements to quantize a neural network with machine learning

A method for classifying an object includes applying multiple confidence values to multiple objects. The method also includes determining a metric based on the multiple confidence values. The method further includes determining a classification of a first object from the multiple objects based on a knowledge-graph when the metric is above a threshold.

Knowledge-graph biased classification for data

A method of quantizing a floating point machine learning network to obtain a fixed point machine learning network using a quantizer may include selecting at least one moment of an input distribution of the floating point machine learning network. The method may also include determining quantizer parameters for quantizing values of the floating point machine learning network based at least in part on the at least one selected moment of the input distribution of the floating point machine learning network to obtain corresponding values of the fixed point machine learning network.

Fixed point neural network based on floating point neural network quantization

A scanning device generally produces an image having a uniform resolution throughout a target region. To improve radar scanning/lidar scanning, an efficient scan approach to enable a radar device/lidar device to adoptively perform a scan of a target region based on interested regions and/or an adjustable resolution. The apparatus may be a scanning device for scanning. The apparatus performs a first scan over a target region to obtain a plurality of first scan samples at a plurality of locations within the target region. The apparatus generates a saliency map of the target region based on signal intensities of the plurality of first scan samples. The apparatus determines a salient region within the target region based on the saliency map. The apparatus performs at least one second scan over the salient region to obtain at least one second scan sample in the salient region.

Saliency based beam-forming for object detection

In an embodiment, a vehicle controller determines a predicted driving performance level of a vehicle driving control entity (VDCE) while the vehicle is controlled by a different VDCE. Driving control is transitioned to the VDCE, after which an actual driving performance level of the VDCE is monitored. The vehicle controller determines whether to transition driving control away from the VDCE based on the actual driving performance level. In another embodiment, after transition driving control to a VDCE, a period of heightened scrutiny used to evaluate the actual driving performance level of the VDCE specifically after the transition.

Managing vehicle driving control entity transitions of an autonomous vehicle based on an evaluation of performance criteria

In general, techniques are described for maintaining occupant awareness in vehicles. A device configured to maintain occupant awareness in a vehicle comprising: a processor and a display may be configured to perform the techniques. The processor may determine a location at which an occupant is gazing, and generate, when the determined location indicates that the occupant is not focused on a direction in which the vehicle is traveling, one or more contextual images capable of assisting the occupant in maintaining awareness of a context in which the vehicle is currently operating. The display may present, based on the determined location, the one or more contextual images proximate to the determined position within the cabin of the vehicle to assist the occupant in assuming control of the vehicle when the vehicle is no longer able to autonomously control the operation of the vehicle.

Maintaining occupant awareness in vehicles

Methods, systems, and devices are described for identifying noisy regions in a skin conductance signal. The signal is divided into a plurality of windows. Two or more features of the signal within a first window are computed. At least one of the two or more features being in a frequency domain. At least two of the features are combined to obtain at least a first metric. The first metric is compared to a corresponding threshold. The first window is identified as a noisy region of the skin conductance signal based on the comparison.

Processing of skin conductance signals to mitigate noise and detect signal features

Certain aspects of the present disclosure relate to a method for compressed sensing (CS). The CS is a signal processing concept wherein significantly fewer sensor measurements than that suggested by Shannon/Nyquist sampling theorem can be used to recover signals with arbitrarily fine resolution. In this disclosure, the CS framework is applied for sensor signal processing in order to support low power robust sensors and reliable communication in Body Area Networks (BANs) for healthcare and fitness applications.

Method and apparatus for distributed processing for wireless sensors

Computing a non-linear function &#x192;(x) in hardware or embedded systems can be complex and resource intensive. In one or more aspects of the disclosure, a method, a computer-readable medium, and an apparatus are provided for computing a non-linear function &#x192;(x) accurately and efficiently in hardware using look-up tables (LUTs) and interpolation or extrapolation. The apparatus may be a processor. The processor computes a non-linear function &#x192;(x) for an input variable x, where &#x192;(x)=g(y(x),z(x)). The processor determines an integer n by determining a position of a most significant bit (MSB) of an input variable x. In addition, the processor determines a value for y(x) based on a first look-up table and the determined integer n. Also, the processor determines a value for z(x) based on n and the input variable x, and based on a second look-up table. Further, the processor computes &#x192;(x) based on the determined values for y(x) and z(x).

Approximation of non-linear functions in fixed point using look-up tables

An intelligent camera network cooperatively acquires images over a wireless network. The network automatically captures images based on a trigger. The trigger may include messages from other image capturing devices. A first image capture devices is triggered to acquire an image based on a message from at least one other image capture device.

Managing crowd sourced photography in a wireless network

Certain aspects of the present disclosure relate to a method for quantizing signals and reconstructing signals, and/or encoding or decoding data for storage or transmission. Points of a signal may be determined as local extrema or points where an absolute rise of the signal is greater than a threshold. The tread and value of the points may be quantized, and certain of the quantizations may be discarded before the quantizations are transmitted. After being received, the signal may be reconstructed from the quantizations using an iterative process.

Method and apparatus for processing and reconstructing data

A method of blink and averted gaze avoidance with a camera includes detecting an averted gaze of a subject and/or one or more closed eyes of the subject in response to receiving an input to actuate a camera shutter. The method also includes scheduling actuation of the camera shutter to a future estimated time period to capture an image of the subject when a gaze direction of the subject is centered on the camera and/or both eyes of the subject are open.

Blink and averted gaze avoidance in photographic images

A method of frequency discrimination associated with the Doppler effect is presented. The method includes mapping a first signal to a first plurality of frequency bins and a second signal to a second plurality of frequency bins. The first signal and the second signal corresponding to different times. The method also includes firing a first plurality of neurons based on contents of the first plurality of frequency bins and firing a second plurality of neurons based on contents of the second plurality of frequency bins.

Doppler effect processing in a neural network model

Growing community of inventors

San Diego, CA, United States of America

Somdeb Majumdar