公开(公告)号：US20190258928A1

公开(公告)日：2019-08-22

申请号：US16280065

申请日：2019-02-20

Applicant: Sony Corporation

Inventor： Javier Alonso Garcia ,  Fabien Cardinaux ,  Thomas Kemp ,  Stephen Tiedemann ,  Stefan Uhlich ,  Kazuki Yoshiyama

IPC: G06N3/08 ,  G06N5/04

Abstract: A computer-implemented method of generating a derived artificial neural network (ANN) from a base ANN comprises initialising a set of parameters of the derived ANN in dependence upon parameters of the base ANN; inferring a set of output data from a set of input data using the base ANN; quantising the set of output data; and training the derived ANN using training data comprising the set of input data and the quantised set of output data.

公开(公告)号：US12072951B2

公开(公告)日：2024-08-27

申请号：US15903290

申请日：2018-02-23

Applicant: SONY CORPORATION

Inventor： Fabien Cardinaux ,  Stefan Uhlich ,  Thomas Kemp ,  Javier Alonso Garcia ,  Kazuki Yoshiyama

IPC: G06F17/16 ,  G06N3/045 ,  G06N3/063 ,  G06N3/08

CPC classification number: G06F17/16 ,  G06N3/08 ,  G06N3/045 ,  G06N3/063

Abstract: An apparatus comprising circuitry that implements an artificial neural network training algorithm that uses weight tying.

公开(公告)号：US12045715B2

公开(公告)日：2024-07-23

申请号：US16244183

申请日：2019-01-10

Applicant: Sony Corporation

Inventor： Javier Alonso Garcia ,  Fabien Cardinaux ,  Kazuki Yoshiyama ,  Thomas Kemp ,  Stephen Tiedemann ,  Stefan Uhlich

IPC: G06N3/08 ,  G06N3/084 ,  G06N5/04

CPC classification number: G06N3/08 ,  G06N3/084 ,  G06N5/04

Abstract: A computer-implemented method of training an artificial neural network (ANN) by generating a first learned parameter for use in normalising input data values during a subsequent inference phase of the trained ANN. The method includes, for each of a series of batches of training data values, deriving a batch variance of the batch of training data values and a running variance of all training data values already processed in the training phase; generating an approximation of a current value of the first learned parameter so that a first scaling factor dependent upon the approximation of the first learned parameter and the running variance, is constrained to be equal to a power of two; and normalizing the batch of input data values by a second scaling factor dependent upon the approximation of the current value of the first learned parameter and the batch variance.

公开(公告)号：US20190258931A1

公开(公告)日：2019-08-22

申请号：US16280059

申请日：2019-02-20

Applicant: Sony Corporation

Inventor： Javier Alonso Garcia ,  Fabien Cardinaux ,  Kazuki Yoshiyama ,  Thomas Kemp ,  Stephen Tiedemann ,  Stefan Uhlich

IPC: G06N3/08 ,  G06N3/04 ,  G06F17/18

Abstract: A computer-implemented method of generating a modified artificial neural network (ANN) from a base ANN having an ordered series of two or more successive layers of neurons, each layer passing data signals to the next layer in the ordered series, the neurons of each layer processing the data signals received from the preceding layer according to an activation function and weights for that layer comprises: detecting the data signals for a first position and a second position in the ordered series of layers of neurons; generating the modified ANN from the base ANN by providing an introduced layer of neurons to provide processing between the first position and the second position with respect to the ordered series of layers of neurons of the base ANN; deriving an initial approximation of at least a set of weights for the introduced layer using a least squares approximation from the data signals detected for the first position and a second position; and processing training data using the modified ANN to train the modified ANN including training the weights of the introduced layer from their initial approximation.

公开(公告)号：US20190220741A1

公开(公告)日：2019-07-18

申请号：US16244183

申请日：2019-01-10

Applicant: Sony Corporation

Inventor： Javier Alonso Garcia ,  Fabien Cardinaux ,  Kazuki Yoshiyama ,  Thomas Kemp ,  Stephen Tiedemann ,  Stefan Uhlich

IPC: G06N3/08 ,  G06N5/04

CPC classification number: G06N3/08 ,  G06N3/084 ,  G06N5/04

Abstract: A computer-implemented method of training an artificial neural network (ANN) by generating a first learned parameter for use in normalising input data values during a subsequent inference phase of the trained ANN. The method includes, for each of a series of batches of training data values, deriving a batch variance of the batch of training data values and a running variance of all training data values already processed in the training phase; generating an approximation of a current value of the first learned parameter so that a first scaling factor dependent upon the approximation of the first learned parameter and the running variance, is constrained to be equal to a power of two; and normalizing the batch of input data values by a second scaling factor dependent upon the approximation of the current value of the first learned parameter and the batch variance.