开源软件名称(OpenSource Name):lonl/CDBN开源软件地址(OpenSource Url):https://github.com/lonl/CDBN开源编程语言(OpenSource Language):MATLAB 72.0%开源软件介绍(OpenSource Introduction):Convolutional Deep Belief Networks with 'MATLAB','MEX','CUDA' versionsThis program is an implementation of Convolutional Deep Belief Networks. In this code, the binary and Gaussian visable types are both supported. In addition, CUDA acceleration is also included. We provide some demo programs to show the usage of the code. Requirement
Build
Run the program
ExperimentsWe have conducted classification experiments with 'Convolutional Deep Belief Networks', 'Deep Belief Networks', and 'Directed Softmax' in mnist data (2000 train data & 2000 test data). The detail parameters of these three ways can be found in code. The comparison results (accuracy) are as follows: No noise added in test data: CDBN: 95.1% DBN: 91.5% Softmax: 87.7% 10% noise added in test data: CDBN: 92.8% DBN: 86.7% Softmax: 83.2% 20% noise added in test data: CDBN: 84.4% DBN: 60.1% Softmax: 74.7% Note
or you can change the computation method in the layer defination, for example, you can add above lines to 'DemoCDBN_Binary_2D.m' at layer 1's defination as:
ConnectionIf you have any problem, or you have some suggestions for this code, please contact me: hanpc839874404@163.com, thank you very much! |
2023-10-27
2022-08-15
2022-08-17
2022-09-23
2022-08-13
请发表评论