Back to results
US 2019/0294972 A1Published application
Representing a neural network utilizing paths within the network to improve a performance of the neural network
AssigneeNVIDIA Corporation
InventorsAlexander Keller, Gonçalo Felipe Torcato Mordido, Noah Jonathan Gamboa, Matthijs Jules Van Keirsbilck
Life of this patent
Estimating the term…
Abstract
Artificial neural networks (ANNs) are computing systems that imitate a human brain by learning to perform tasks by considering examples. By representing an artificial neural network utilizing individual paths each connecting an input of the ANN to an output of the ANN, a complexity of the ANN may be reduced, and the ANN may be trained and implemented in a much faster manner when compared to an implementation using fully connected ANN graphs.
Claims28
A method comprising:
- creating an artificial neural network (ANN) represented by a plurality of paths each connecting at least one input of the ANN to at least one output of the ANN,
- where each of the plurality of paths includes a plurality of vertices each representing a neural unit within the ANN, and a plurality of edges each representing a weighted connection within the ANN.
A method comprising:
- compressing an artificial neural network (ANN) by subsampling a plurality of edges each representing a weighted connection within the ANN.
A method comprising:
- performing network normalization by:
- normalizing a plurality of weights of a first layer of an artificial neural network (ANN) with a linear factor;
- and
- propagating the normalized plurality of weights to successive layers of the ANN by multiplying weights of the successive layers by the linear factor;
- wherein a last layer of the ANN stores resulting weights to scale outputs of the ANN.
2 more elements
Dependencies are inferred from the claim text.