This repository has been archived by the owner on Jan 3, 2024. It is now read-only.
Releases: matiasvlevi/Dann
Releases · matiasvlevi/Dann
v2.2.2d
Added dev environment setup
- Running
devsetup.bat
will install node modules (or update them) frompackage.json
and setup a test environment for the browser & nodejs
test.bat
runs the tests/nodeTests.js
file.
test.htm
uses the tests/browserTests.js
file.
fix
mutateRandom();
probability argument is now optional
v2.2.2c
Patch
- fixed the
outputActivation();
problem with the browser
v2.2.2b
Patch
- patched the
Dann error: x,y arguments exceed the matrix dimensions.
when logging a dann model with{detials:true}
v2.2.2
Segment Update
- the initial
dann.js
source file is now segmented in/src
. - Grunt is used to concat & minify. Built files are located in
/build
Other
poolFuncs
is nowpoolfuncs
- A lot of global functions are now static Layer functions for cleaner code.
v2.2.1b
README changes
v2.2.1
Added
- functions for the matrix object.
- error handling for new & pre-existing matrix functions.
New functions
insert( value , x , y );
This function sets a specific value in the matrix.
- value
The value to be inserted into the specified coordinates in the matrix
- x
Column index
- y
Row index
initiate( value );
This function resets the matrix, all values are set to the same value.
- value (optional)
The value to be set through out the matrix. Is set to 0 by default
v2.2.0
Dann.save(); changes
- changed the way the csv train report is formatted.
Dann.load(); changes
- you can now add a callback as the second argument for nodejs.
Here is an example:
const nn = new Dann(1,1);
nn.load('sampleA', function () {
console.log('finished loading');
});
v2.1.9d
Pooling layer v2.1.9d
pooling layers now can take an input array that can be arranged as a non square matrix. Width & Height arguments are optional, if not specified, the layer will arrange the array as a square matrix.
const layer = new Layer( type , inputSize , sampleSize , stride , width , height );
Example:
const layer = new Layer('maxpool', 12, 2, 1, 3, 4);
// Array length of 12 (can be arranged in a 3x4 matrix)
let input = [
1,2,3,
4,5,6,
7,8,9,
10,11,12
];
let output = layer.feed(input);
// output = [ 5, 6, 8, 9, 11, 12];
v2.1.9c
- getPoolOutputLength(); fix
v2.1.9b
- pooling layer input undefined rows matrix fix