FIELD: information technology.
SUBSTANCE: group of inventions relates to neural networks. Method comprises obtaining a trained convolutional neural network comprising one or more convolutional levels, wherein each of the one or more convolutional levels comprises a plurality of filters with known filter parameters; preliminary calculation of multiple use factor for each of one or more convolutional levels based on known filter parameters of trained convolutional neural network; receiving input data in a trained convolutional neural network; calculating the output of each of the one or more convolutional levels using the Winograd convolution operator based on the pre-calculated reusability coefficient and the input data; and determining output data of the trained convolutional network based on the output of each of the one or more convolutional levels.
EFFECT: technical result is providing fast calculation of convolutional neural network.
17 cl, 4 dwg
| Title | Year | Author | Number |
|---|---|---|---|
| METHOD AND SERVER FOR CONVERTING TEXT TO SPEECH | 2020 |
|
RU2775821C2 |
| REPRODUCING AUGMENTATION OF IMAGE DATA | 2018 |
|
RU2716322C2 |
| METHOD OF RECOGNIZING SPEECH EMOTIONS USING 3D CONVOLUTIONAL NEURAL NETWORK | 2023 |
|
RU2816680C1 |
| IMAGE PROCESSING SYSTEM AND METHOD USING DEEP NEURAL NETWORKS | 2018 |
|
RU2743931C1 |
| CHARACTER RECOGNITION USING A HIERARCHICAL CLASSIFICATION | 2018 |
|
RU2693916C1 |
| METHODS FOR RECONSTRUCTION OF DEPTH MAP AND ELECTRONIC COMPUTER DEVICE FOR THEIR IMPLEMENTATION | 2020 |
|
RU2745010C1 |
| METHOD AND ELECTRONIC DEVICE FOR DETECTING THREE-DIMENSIONAL OBJECTS USING NEURAL NETWORKS | 2021 |
|
RU2776814C1 |
| SPLICING SITES CLASSIFICATION BASED ON DEEP LEARNING | 2018 |
|
RU2780442C2 |
| SUPERRESOLUTION OF AN IMAGE IMITATING HIGH DETAIL BASED ON AN OPTICAL SYSTEM, PERFORMED ON A MOBILE DEVICE HAVING LIMITED RESOURCES, AND A MOBILE DEVICE WHICH IMPLEMENTS | 2018 |
|
RU2697928C1 |
| CONVOLUTIONAL NEURAL NETWORK BASED ON OCTREE | 2018 |
|
RU2767162C2 |
