Qualcomm Incorporated

Other

Qualcomm Innovation Center, Inc.

Christopher G Lott

Rashid Ahmed Akbar Attar

Donna Ghosh

Rashid Ahmed Attar

Peter J Black

Joseph Binamira Soriaga

Linhai He

Jun Hu

Jing X Sun

Naga Bhushan

Jun Ma

Ravindra Manohar Patwardhan

Jean Put Ling Au

George Cherian

Peerapol Tinnakornsrisuphap

Ramin Rezaiifar

Radhika Dilip Gowaikar

Yang Yang

Yu-Cheun Jou

Serafin Diaz Spindola

Jin Won Lee

Rexford Alan Hill

Eduardo Esteves

Vikram Makam Gupta

Rohit Kapoor

Daniel Fred Filipovic

Qingxin Chen

Kambiz Azarian Yazdi

Kiran Kumar Somasundaram

Venkata Ramanan Venkatachalam Jayaraman

Gerhard Reitmayr

Francis Ming-Meng Ngai

Amin Ansari

Yen-Chi Lee

Aaron Douglass Lamb

Casimir Matthew Wierzynski

Bardia Fallah Behabadi

Aliakbar Aghamohammadi

Michael Goldfarb

Rashid Ahmed A Attar

Mingu Lee

Roland Memisevic

Wonseok Jeon

Anusha Lalitha

Kambiz Azarian-Yazdi

Juan Montojo

Arvind Vardarajan Santhanam

Sitaramanjaneyulu Kanamarlapudi

Jilei Hou

Edward Harrison Teague

Amir Farajidana

Xiliang Luo

Won-Joon Choi

Mingxi Fan

Mark Bapst

Weiliang Zeng

Vinod Viswanatha Menon

Chaitanya Mannemala

Chandrasekhar Therazhandur Sundarraman

Sarah Paige Gibson

Saurav Agarwal

Shayegan Omidshafiei

Ming Yan

Rohan Salvi

Ravindra Mahendrakumar Garach

Dinesh K Garg

Nikolai Konrad Nepomuceno Leung

Sunit Bhatia

Piero Zappi

Chunhao Chen

Rashmi Keshava Iyengar

Gopal Chillariga

Kang Wang

Jean Au

Amr Mamoun Martini

Shweta Nandakumar

Guan-Ming Su

Sudheer Konda

Rashid A Arrar

Certain aspects provide techniques and apparatus for executing queries in a computing system using machine learning models. An example method generally includes receiving a plan to satisfy a request in the computing system and event log data associated with execution of the plan. The plan generally specifies a first plurality of actions to be performed by the computing system at a first level of granularity. Using a plan refinement machine learning model, a refined plan is generated when the event log data indicates that execution of the generated plan results in one or more execution errors and the one or more execution errors are solvable. Generally, the refined plan specifies a second plurality of actions to be performed by the computing system at a second level of granularity, the second level of granularity being finer than the first level of granularity.

Executing queries in computing systems using execution plans generated by generative artificial intelligence models

A method for optimizing the compilation of a machine learning model to be executed on target edge devices is provided. Compute nodes of a plurality of compute nodes are allocated to a compiler optimization process for a compiler of said machine learning model. The machine learning model has a compute graph representation having nodes that are kernel operators necessary to execute the machine learning model and edges that connect said kernel operators to define precedence constraints. A round of optimization is scheduled for the process amongst the allocated compute nodes. At each allocated compute node a sequencing and scheduling solution is applied per round to obtain a performance metric for the machine learning model. From each compute node the performance metric is received and a solution that has the best performance metric is identified and implemented for execution of the machine learning model on the target edge devices.

Distributed machine learning compiler optimization

A method performed by a computing device includes determining a partition for depth-first processing by a multi-layer artificial neural network (ANN) of the computing device. The computing device comprising a processor, on-chip memory, and off-chip memory. The first partition determined based on an amount of on-chip memory used by the first partition, an available amount of on-chip memory, and a size of a write back to the off-chip memory. The method also includes processing, at the device via the multi-layer ANN, an input, using the depth-first processing in accordance with the partition.

Depth-first deep convolutional neural network inference

Certain aspects of the present disclosure provide techniques and apparatus for generating a response to a query input in a generative artificial intelligence model. An example method generally includes receiving a plurality of sets of tokens generated based on an input prompt and a first generative artificial intelligence model, each set of tokens in the plurality of sets of tokens corresponding to a candidate response to the input prompt; selecting, using a second generative artificial intelligence model and recursive adjustment of a target distribution associated with the received plurality of sets of tokens, a set of tokens from the plurality of sets of tokens; and outputting the selected set of tokens as a response to the input prompt.

Speculative decoding in autoregressive generative artificial intelligence models

A method of exploiting activation sparsity in deep neural networks is described. The method includes retrieving an activation tensor and a weight tensor where the activation tensor is a sparse activation tensor. The method also includes generating a compressed activation tensor comprising non-zero activations of the activation tensor, where the compressed activation tensor has fewer columns than the activation tensor. The method further includes processing the compressed activation tensor and the weight tensor to generate an output tensor.

Exploiting activation sparsity in deep neural networks

Various embodiments include methods and devices for implementing automatic grammar augmentation for improving voice command recognition accuracy in systems with a small footprint acoustic model. Alternative expressions that may capture acoustic model decoding variations may be added to a grammar set. An acoustic model-specific statistical pronunciation dictionary may be derived by running the acoustic model through a large general speech dataset and constructing a command-specific candidate set containing potential grammar expressions. Greedy based and cross-entropy-method (CEM) based algorithms may be utilized to search the candidate set for augmentations with improved recognition accuracy.

Automatic grammar augmentation for robust voice command recognition

A method for defining a sensor model includes determining a probability of obtaining a measurement from multiple potential causes in a field of view of a sensor modeled based on a stochastic map. The stochastic map includes a mean occupancy level for each voxel in the stochastic map and a variance of the mean occupancy level for each pixel. The method also includes determining a probability of obtaining an image based on the determined probability of obtaining the measurement. The method further includes planning an action for a robot, comprising the sensor, based on the probability of obtaining the image.

Stochastic map-aware stereo vision sensor model

A method substantially simultaneously plans a path and maps an environment by a robot. The method determines a mean of an occupancy level for a location in a map. The method also includes determining a probability distribution function (PDF) of the occupancy level. The method further includes calculating a cost function based on the PDF. Finally, the method includes simultaneously planning the path and mapping the environment based on the cost function.

Simultaneous mapping and planning by a robot

Embodiments disclosed herein relate to carrier allocation and management in multi-carrier communication systems. In some embodiments, the number of carriers assigned to an access terminal on a forward link may be determined by an access network, and the number of carriers assigned to the access terminal on a reverse link may be based on a cooperative process between the access terminal and the access network. In other embodiments, the number of carriers assigned to the access terminal on the reverse link may also be determined by the access network, e.g., in relation to the scheduling information received from the access terminal.

Method and apparatus for carrier allocation and management in multi-carrier communication systems

Some implementations provide a method for prioritizing reverse link packets so that particular reverse link packets are transmitted with a reduced latency in order to improve spectral efficiency of the forward link. The method also optionally includes weighting packets according to their determined lengths, and transmitting packets based on the determined weights. In one implementation, the method includes sorting packets at least according to one of size, type and nature of the data in the packets. Accordingly, sorting of the packets includes at least one of determining the size and the type of each packet. In one implementation, sorting the packets includes deep-packet inspection to determine a respective priority value for each packet.

Reverse link control to reduce forward link latency

Techniques for the reuse of airlink resources in a cellular network are disclosed. A base station controller may gather information about load levels at adjacent sectors of the cellular network. The base station controller may generate and distribute a reuse reference to base stations serving mobile devices in the adjacent sectors. Alternatively, the base station controller may provide the load level information to base stations serving the adjacent sectors and the base stations may determine a sector reuse pattern. The reuse reference or reuse pattern may be adapted to sector load conditions and may designate any combination of carriers, time slots, and power levels for data transmission in a sector. The base stations may schedule data transmission to mobile devices in their respective sectors according to the reuse reference or reuse pattern. The base stations may modify their resource usage in response to changes in sector load.

Adaptive semi-static interference avoidance in cellular networks

Aspects of the present disclosure are directed to improving maximum transmit power in multi-carrier reverse-link transmission. In one aspect, a method of carrier management for a multi-carrier reverse link transmission is disclosed. A method can include transmitting a reverse link signal on a plurality of carriers, and the reverse link signal including payload data and overhead data. A method can funnel payload data onto a first carrier of the plurality of carriers, while maintaining transmission of the overhead data on all the carriers of the plurality of carriers. Other aspects, embodiments, and features are also claimed and described.

Apparatus and methods for enhanced maximum power in multicarrier wireless communications

Growing community of inventors

San Diego, CA, United States of America

Christopher G Lott