Intel Corporation

Amazon Technologies, Inc.

University of Illinois

Oguz Hasan Elibol

Grace M Credo

Jonathan S Daniels

Xing Su

Anil Thomas

Gokce Keskin

Joydeep Ray

Elmoustapha Ould-Ahmed-Vall

Prasoonkumar Surti

Rajesh Poornachandran

Ben J Ashbaugh

Chandrasekaran Sakthivel

Jayaram Bobba

Kumar Balasubramanian

Jerin C Justin

Pradeep Ramani

Guozhong Zhuang

Bhavani Subramanian

Jing Huang

Xiaoming Cui

Ting Gong

Timothy B Costa

Rama Harihara

Roberto Barra Chicote

Noureddine Tayebi

Yixiong Meng

Itay Teller

Travis T Schluessler

Eriko Nurvitadhi

Yi-Shao Liu

Tomer Bar-On

Abhishek Venkatesh

Devan Burke

Andrew Jay Roths

James G Droppo

Amit Bleiweiss

John G Gierach

Rashid Bashir

Onur Can Akkaya

Handong Li

Kai Wu

Esha John

Dhawal Srivastava

Zeya Chen

Hongda Mao

Bobby Reddy, Jr

Ryan M Field

Varvara Kollia

Jie Liang

Udhgee Murugesan

Mansour Ahmadi

Eunjung Han

Ruirui Li

Yuguang Yang

Mohamed G Mahmoud

Andreas Stolcke

Nisarga Naik

Madoo Varma

Drew Hall

Ashraf Alam

Demir Akin

Roger Hau-Bin Cheng

Donald E Bergstrom

Stephane L Smith

Zhenning Tan

Grzegorz Beringer

Chia-Hao Chang

Pruthvi Gowda Thorehosur Appajigowda

Nishant Patel

Huma Abidi

Jaikrishnan Menon

Kevin Kwanghoon Ha

Kian Jamali Abianeh

Techniques for performing voice-based user recognition are described. When a device receives audio data corresponding to a spoken user input, the device may generate spoken user input embedding data representing speech characteristics of the spoken user input. The device may also identify user embedding data representing speech characteristics of a user known to the device. A machine learning (ML) model may process the spoken user input embedding data to generate reduced spoken user input embedding data including a reduced number of dimensions, where the reduced number of dimensions are functions of the higher number of dimensions of the spoken user input embedding data. Moreover, the ML model may process the user embedding data to generate reduced user embedding data including a reduced number of dimensions, where the reduced number of dimensions are functions of the higher number of dimensions of the user embedding data and are tuned to characteristics of users of the device (as opposed to a representation of all possible users). A comparison of the reduced spoken user input embedding data and the reduced user embedding data may be used to determine whether the user spoke the user input.

Voice-based user recognition

An apparatus to facilitate compute optimization is disclosed. The apparatus includes a at least one processor to perform operations to implement a neural network and compute logic to accelerate neural network computations.

Compute optimization mechanism for deep neural networks

Embodiments are generally directed to compression in machine learning and deep learning processing. An embodiment of an apparatus for compression of untyped data includes a graphical processing unit (GPU) including a data compression pipeline, the data compression pipeline including a data port coupled with one or more shader cores, wherein the data port is to allow transfer of untyped data without format conversion, and a 3D compression/decompression unit to provide for compression of untyped data to be stored to a memory subsystem and decompression of untyped data from the memory subsystem.

Compression of machine learning models utilizing pseudo-labeled data training

Techniques for determining whether audio is machine-outputted or non-machine-outputted are described. A device may receive audio, may process the audio to determine audio data including audio features corresponding to the audio, and may process the audio data to determine audio embedding data. The device may process the audio embedding data to determine whether the audio is machine-outputted or non-machine-outputted. In response to determining that the audio is machine-outputted, then the audio may be discarded or not processed further. Alternatively, in response to determining that the audio is non-machine-outputted (e.g., live speech from a user), then the audio may be processed further (e.g., using ASR processing).

Detecting machine-outputted audio

Dynamic assignment of down sampling intervals for data stream processing

An embodiment of a semiconductor package apparatus may include technology to process one or more vectors with a sum of squares operation with a layer of a multi-layer neural network, and determine a fixed-point approximation for the sum of squares operation. Other embodiments are disclosed and claimed.

Iterative normalization for machine learning applications

A speech-processing system receives input data representing text. A first encoder processes segments of the text to determine embedding data representing the text, and a second encoder processes corresponding audio data to determine prosodic data corresponding to the text. The embedding and prosodic data is processed to create output data including a representation of speech corresponding to the text and prosody.

Synthetic speech processing

Policy-based system interface for a real-time autonomous system

Compression in machine learning and deep learning processing

Described herein is a device comprising a plurality of first reaction electrodes arranged in an array, the plurality of first reaction electrodes configured to be exposed to a solution and having a capacitance; first circuitry configured to controllably connect the plurality of first reaction electrodes to a bias source and controllably disconnect the plurality of first reaction electrodes from the bias source; and second circuitry configured to measure a rate of charging or discharging of the capacitance. Also described herein is a method of using this device to sequence DNA.

Device and method for detecting redox reactions in solution

An apparatus including a circuit substrate including a contact in a metal layer; and a transducer including a first electrode deposited on and coupled to a sidewall of the contact and a second electrode coupled to a conductor through which voltage can be applied, wherein the second electrode includes a profile aligned to the sidewall of the contact and separated from the first electrode by a gap. A method including forming a transducer adjacent a contact in a metal layer on a substrate, the transducer including a first electrode disposed on a sidewall of the contact and a second electrode coupled to a conductor through which voltage can be applied, wherein the second electrode includes a profile aligned to the sidewall of the contact and separated from the first electrode by a gap.

Self aligned and scalable nanogap post processing for DNA sequencing

An analog frontend (AFE) interface is dynamically programmable based on a determined operating state. The AFE includes hardware to interface with multiple different sensors. The AFE includes analog processing hardware that can select input data from one of the multiple sensors. The analog processing hardware is coupled to a processor that computes features from the sensor, where the features represent selected operating condition information of the AFE for the sensor. The processor is to determine one of multiple discrete operating states of the AFE for the sensor based on the computed features and dynamically adjust operation of the AFE to interface with the sensor based on the determined operating state. Dynamically adjusting the operation of the AFE includes controlling a configuration of the AFE that controls how the AFE receives the input sensor data from the sensor.

Low-power context-aware control for analog frontend

Disclosed herein is a method comprising: depositing a second electrode of each of a plurality of electrode pairs onto a substrate, through an opening of one or more resist layers; depositing a strip of a sacrificial layer directly on the second electrode through the same opening of the one or more resist layer; depositing a first electrode of each of the plurality of electrode pairs directly on the strip of the sacrificial layer through the same opening of the one or more resist layer; and forming a nanogap channel by removing the strip of the sacrificial layer; wherein the strip of the sacrificial layer is sandwiched between and in direct contact with the first electrode and the second electrode before the strip is removed, and wherein at least a portion of the first electrode directly faces at least a portion of the second electrode.

Self-aligned nanogap fabrication

Various embodiments provide devices, methods, and systems for high throughput biomolecule detection using transducer arrays. In one embodiment, a transducer array made up of transducer elements may be used to detect byproducts from chemical reactions that involve redox genic tags. Each transducer element may include at least a reaction chamber and a fingerprinting region, configured to flow a fluid from the reaction chamber through the fingerprinting region. The reaction chamber can include a molecule attachment region and the fingerprinting region can include at least one set of electrodes separated by a nanogap for conducting redox cycling reactions. In embodiments, by flowing the chamber content obtained from a reaction of a latent redox tagged probe molecule, a catalyst, and a target molecule in the reaction chamber through the fingerprinting region, the redox cycling reactions can be detected to identify redox-tagged biomolecules.

High throughput biochemical detection using single molecule fingerprinting arrays

Embodiments of the invention provide transducers capable of functioning as electronic sensors and redox cycling sensors. Transducers comprise two electrodes separated by a nanogap. Molecular binding regions proximate to and within the nanogap are provided. Methods of fabricating nanogap transducers and arrays of nanogap transducers are also provided. Arrays of individually addressable nanogap transducers can be disposed on integrated circuit chips and operably coupled to the integrated circuit chip.

Nanogap transducers with selective surface immobilization sites

Growing community of inventors

Sunnyvale, CA, United States of America

Oguz Hasan Elibol