Exxonmobil Upstream Research Company

Google Inc.

Exxonmobil Research and Engineering Company

Niranjan A Subrahmanya

Wei D Liu

Aleks Kracun

Aishanee Shah

Ignacio Lopez Moreno

Randy C Tolman

Krishnan Kumaran

Pavlin B Entchev

Renzo Moises Angeles Boza

Huseyin Denli

Hyun Jin Park

Kuang-Hung Liu

Amr S El-Bakry

Peng Xu

Thomas A Badgwell

Diego A Hernandez

Michael H Kovalski

Pai Zhu

D Braden Fitz-Gerald

Joseph K Bjerkseth

Renzo M Angeles Boza

Teacher-student learning can be used to train a keyword spotting (KWS) model using augmented training instance(s). Various implementations include aggressively augmenting (e.g., using spectral augmentation) base audio data to generate augmented audio data, where one or more portions of the base instance of audio data can be masked in the augmented instance of audio data (e.g., one or more time frames can be masked, one or more frequencies can be masked, etc.). Many implementations include processing augmented audio data using a KWS teacher model to generate a soft label, and processing the augmented audio data using a KWS student model to generate predicted output. One or more portions of the KWS student model can be updated based on a comparison of the soft label and the generated predicted output.

Noisy student teacher training for robust keyword spotting

Techniques are described herein for improving performance of machine learning model(s) and thresholds utilized in determining whether automated assistant function(s) are to be initiated. A method includes: receiving, via one or more microphones of a client device, audio data that captures a spoken utterance of a user; processing the audio data using a machine learning model to generate a predicted output that indicates a probability of one or more hotwords being present in the audio data; determining that the predicted output satisfies a secondary threshold that is less indicative of the one or more hotwords being present in the audio data than is a primary threshold; in response to determining that the predicted output satisfies the secondary threshold, prompting the user to indicate whether or not the spoken utterance includes a hotword; receiving, from the user, a response to the prompting; and adjusting the primary threshold based on the response.

User mediation for hotword/keyword detection

A method to automatically interpret a subsurface feature within geophysical data, the method including: storing, in a computer memory, geophysical data obtained from a survey of a subsurface region; and extracting, with a computer, a feature probability volume by processing the geophysical data with one or more fully convolutional neural networks, which are trained to relate the geophysical data to at least one subsurface feature, wherein the extracting includes fusing together outputs of the one or more fully convolutional neural networks.

Automated seismic interpretation using fully convolutional neural networks

A method including: storing, in a computer memory, geophysical data obtained from a survey of a subsurface region; and extracting, with a computer, a subsurface physical property model by processing the geophysical data with one or more convolutional neural networks, which are trained to relate the geophysical data to at least one subsurface physical property consistent with geological prior information.

Geophysical inversion with convolutional neural networks

Systems and methods are provided for using a Deep Reinforcement Learning (DRL) agent to provide adaptive tuning of process controllers, such as Proportional-Integral-Derivative (PID) controllers. The agent can monitor process controller performance, and if unsatisfactory, can attempt to improve it by making incremental changes to the tuning parameters for the process controller. The effect of a tuning change can then be observed by the agent and used to update the agent's process controller tuning policy. It has been unexpectedly discovered that providing adaptive tuning based on incremental changes in tuning parameters, as opposed to making changes independent of current values of the tuning parameters, can provide enhanced or improved control over a controlled variable of a process.

Adaptive PID controller tuning via deep reinforcement learning

A diagnostic apparatus configured to communicate with a well test system comprising a plurality of wells in a field, comprising a receiving component configured to receive a well test data from the well test system, a transmitting component configured to transmit an abnormal well test signal indication, at least one processor configured to communicate with the transmitting component and the receiving component, and a memory coupled to the at least one processor, wherein the memory comprises instructions that when executed by the at least one processor cause the diagnostic apparatus to compare the well test data to one or more well test descriptors stored in memory, correlate the well test data to an abnormal well test result selected based at least in part on the comparison with the one or more well test descriptors stored in the memory, and instruct the transmitting component to transmit an abnormal well test signal indication to a recipient.

Automated well test validation

Methods and apparatus for actuating a downhole tool in wellbore includes acquiring a CCL data set or log from the wellbore that correlates recorded magnetic signals with measured depth, and selects a location within the wellbore for actuation of a wellbore device. The CCL log is then downloaded into an autonomous tool. The tool is programmed to sense collars as a function of time, thereby providing a second CCL log. The autonomous tool also matches sensed collars with physical signature from the first CCL log and then self-actuates the wellbore device at the selected location based upon a correlation of the first and second CCL logs.

Method for automatic control and positioning of autonomous downhole tools

Growing community of inventors

Mountain View, CA, United States of America

Niranjan A Subrahmanya