Activity for dibnn
-
Peter Mills committed [b25553] on Code
Error in test make file for regression problems.
-
Had forgotten to add dib_count2.c (is it even relevant at this point?)...
-
Added some simple statistics for the derivative validation routine
-
- Had another crack at the "hyper-Newton" method. Still doesn't work.
-
One more attempt at the "hyper-Newton" method. Still doesn't work...
-
Testing convolutional NN with mnist dataset. Still doesn't work properly.
-
Need some kind of 'layer index' to make splitting and fusing layers easier.
-
popen does not appear to be either ANSI c99 or c11 compatible...
-
Software is almost completely C99 compliant.
-
Minor fixes: monochrome images can be normalized to lie btw. -1 and 1
-
Switched to monochrome image formats.
-
Initial tests on convolutional NN are promising, but poor...
-
Beginning of test routines for convolutional NN.
-
Adding module name to every error message.
-
Minor edit to make dib_extract work properly.
-
Basic commands for extracting image data seem to work.
-
Serious indexing errors corrected.
-
Finished coding more user friendly convolutional NN.
-
Minor edits.
-
Rearranging stuff and removing redundant, deprecated routines.
-
Have coded the allocation routines for the new, easy-design conv. NN
-
Subroutines for counting... code with complex calling sequence isn't significantly slower...
-
Seem to have gotten the comination-generator I need working...
-
Why can't I get this simple program to work?
-
Swapped raw datasets for their normalized versions.
-
Starting work on more sophisticated method of specifying convolutional NNs.
-
Cleaning up some of the documentation, chiefly in the header files.
-
Formatted the README so it looks better on Sourceforge.
-
Test script now prints out accuracy and U.C. for statistical classification
-
Pattern recognition in a 2-D image using a convolutional NN.
-
Trying to make all the file conversion routines for applying convolutional NN
-
Converting raster images to .vec files.
-
Raster image reading routine seems to work.
-
Raster interface for convolutional neural networks.
-
Changed model_func function template so that it is used to define object class
-
Fixed some errors in the convolutional NN allocation.
-
First step towards getting a working convolutional NN.
-
Changed the uppercase 'NN' to 'DIB' to reflect the new, public name of the
-
Changed the name of the activation function to reflect the common term
-
Adding some missing documentation.
-
Added some annotations to make test routines more user-friendly.
-
Interesting new take on the "class borders" idea...
-
Added GPL v3.0 text.
-
Forgot to remove copyright notieces...
-
Successfully changed the name from 'libnn' to 'dibnn'!
-
Making a commit before name change.
-
Added README file.
-
Moved all the global variables into a structure.
-
"border_class" ML model works and produces reasonable, though not great, results
-
"add_layer" routine works for building up perceptrons and single-layer NNs
-
New paradigm for building neural nets.
-
Starting convolutional implementation.
-
Preallocation works for on NN type. Makes practically no difference.
-
Implementing pre-allocation routines to improve performance.
-
Routines to pre-allocate workspace variables.
-
Created stand-alone "lightweight" neural network library.
1
