Compressing a deep neural network
AssigneeRobert Bosch GmbH
InventorThomas Pfeil
Life of this patent
Estimating the term…
Abstract
A method for compressing a deep neural network. The deep neural network includes a multitude of layers, which are each connected on the input side according to a predefinable sequence to their directly preceding layer of the sequence. The method includes adding a skip connection. Thereafter, a reduction of a resolution of the parameters of the layers follows. A computer program as well as to a device for carrying out the method are also described.
Claims9
A method for compressing a deep neural network, the deep neural network including a multitude of layers, which are each connected on an input side according to a predefinable sequence to their directly preceding layer of the predefinable sequence, parameters being assigned to each layer, the method comprising the following steps:
- adding a skip connection to the deep neural network, the skip connection connecting one of the multitude of layers on the input side to a further layer which is an indirectly preceding layer in the sequence;
- and
- reducing a resolution of the parameters of the layers.
A non-transitory machine-readable memory medium on which is stored a computer program for compressing a deep neural network, the deep neural network including a multitude of layers, which are each connected on an input side according to a predefinable sequence to their directly preceding layer of the predefinable sequence, parameters being assigned to each layer, the computer program, when executed by a computer, causing the computer to perform the following steps:
- adding a skip connection to the deep neural network, the skip connection connecting one of the multitude of layers on the input side to a further layer which is an indirectly preceding layer in the sequence;
- and
- reducing a resolution of the parameters of the layers.
A device for compressing a deep neural network, the deep neural network including a multitude of layers, which are each connected on an input side according to a predefinable sequence to their directly preceding layer of the predefinable sequence, parameters being assigned to each layer, the device configured to:
- add a skip connection to the deep neural network, the skip connection connecting one of the multitude of layers on the input side to a further layer which is an indirectly preceding layer in the sequence;
- and
- reduce a resolution of the parameters of the layers.
Dependencies are inferred from the claim text.