发明名称 |
Simulation of convolutional network behavior and visualizing internal states of a network |
摘要 |
Convolutional networks can be defined by a set of layers being respectively made up by a two-dimensional lattice of neurons. Each layer-with the exception of the last layer-represents a source layer for respectively following target layer. A plurality of neurons of a source layer called a source sub-area respectively share the identical connectivity weight matrix type. Each connectivity weight matrix type is represented by a scalar product of an encoding filter and a decoding filter. For each source layer a source reconstruction image is calculated on the basis of the corresponding encoding filters and the activities of the corresponding source sub-area. For each connectivity weight matrix type, each target sub-area and each target layer the input of the target layer is calculated as a convolution of the source reconstruction image and the decoding filter. For each target layer the activities are calculated by using the non-linear local response function of the neurons of the target layer and the calculated input of the target layer.
|
申请公布号 |
US7236961(B2) |
申请公布日期 |
2007.06.26 |
申请号 |
US20020099364 |
申请日期 |
2002.03.12 |
申请人 |
HONDA RESEARCH INSTITUTE EUROPE GMBH |
发明人 |
EGGERT JULIAN;BAEUML BERTHOLD |
分类号 |
G06F15/00;G06F17/10;G06F17/00;G06F17/14;G06F17/16;G06K9/46;G06N3/00;G06N3/04;G06N3/10 |
主分类号 |
G06F15/00 |
代理机构 |
|
代理人 |
|
主权项 |
|
地址 |
|