公开(公告)号：US20240177048A1

公开(公告)日：2024-05-30

申请号：US18056379

申请日：2022-11-17

Applicant: QUALCOMM Incorporated

Inventor： Weiliang ZENG ,  Christopher G. LOTT ,  Edward H. TEAGUE ,  Yang YANG ,  Joseph Binamira SORIAGA

IPC: G06N20/00

CPC classification number: G06N20/00

Abstract: 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.

公开(公告)号：US20240119301A1

公开(公告)日：2024-04-11

申请号：US18464996

申请日：2023-09-11

Applicant: QUALCOMM Incorporated

Inventor： Wonseok JEON ,  Mukul GAGRANI ,  Weiliang ZENG ,  Edward TEAGUE ,  Burak BARTAN ,  Piero ZAPPI ,  Christopher LOTT

IPC: G06N3/092

CPC classification number: G06N3/092

Abstract: A processor-implemented method includes sampling, according to a priority sampling policy, a set of node priorities from a computation graph. Each node priority of the set of node priorities may be associated with a respective node on the computation graph. Additionally, each node may represent an operation of a task performed by an artificial neural network. The method also includes converting, via a list scheduling function, the node priorities to a schedule that associates each node of the computation graph with a processor of a group of processors of a device associated with the artificial neural network, the schedule associated with a makespan. The method further includes performing the task in accordance with the schedule.

公开(公告)号：US20230164791A1

公开(公告)日：2023-05-25

申请号：US17535201

申请日：2021-11-24

Applicant: QUALCOMM Incorporated

Inventor： Weiliang ZENG ,  Sanaz BARGHI ,  Pouriya SADEGHI ,  Navin Dunichand ANWANI ,  Supratik BHATTACHARJEE ,  Gautham HARIHARAN

IPC: H04W72/04 ,  H04W24/02

CPC classification number: H04W72/042 ,  H04W24/02 ,  H04W72/0446

Abstract: A method of wireless communications by a user equipment (UE) includes decoding information received from a base station via a physical downlink shared channel (PDSCH). The UE determines whether to update channel state feedback based on the decoded information. The channel state feedback is updated based on the determination to generate updated channel state feedback. The updated channel state feedback is transmitted to the base station. The UE is configured to update the channel state feedback without an additional measurement of a channel state information reference signal (CSI-RS).

公开(公告)号：US20230114870A1

公开(公告)日：2023-04-13

申请号：US17498651

申请日：2021-10-11

Applicant: QUALCOMM Incorporated

Inventor： June NAMGOONG ,  Taesang YOO ,  Hyojin LEE ,  Naga BHUSHAN ,  Weiliang ZENG

IPC: G06N3/08 ,  H04W24/10

Abstract: A method of wireless communication by a user equipment (UE) includes receiving different sets of parameters from different sources as input to a receiver neural network. The method also includes receiving, from a base station, a set of target long-term energy values associated with the receiver neural network. The method further includes calculating a scaling factor for each of the different sets of parameters based on the set of target long-term energy values, and separately scaling each of the different sets of parameters based on the scaling factor calculated for that set in order to generate multiple sets of scaled parameters. The method still further includes transmitting the multiple sets of scaled parameters to the receiver neural network.

公开(公告)号：US20210376895A1

公开(公告)日：2021-12-02

申请号：US16888593

申请日：2020-05-29

Applicant: QUALCOMM Incorporated

Inventor： Yisheng XUE ,  Weiliang ZENG ,  Xiaoxia ZHANG ,  Yongbin WEI

IPC: H04B7/06 ,  H04L12/24 ,  H04L1/18 ,  G06N20/00

Abstract: Certain aspects of the present disclosure provide techniques for qualifying machine learning model-based channel state information (CSI) predictions. An example method generally includes receiving, from a network entity, a channel state information (CSI) prediction model for quantized CSI, calculating CSI based on downlink reference signal measurements, generating a quantized CSI difference value based a quantization of a difference between the calculated CSI and CSI predicted based on a CSI prediction model, and reporting, to the network entity, the calculated CSI and the quantized CSI difference value.