piotrbajdek / fortran_ai

artificial intelligence/ machine learning algorithms in Fortran

The repository contains a set of artificial intelligence/ machine learning algorithms in Fortran.

It is available under the .

File	Description
knn_8.f90	k-nearest neighbours regression algorithm
knn_evolutionary.f90	Evolutionary algorithm finding the optimal k value for the k-nearest neighbours method
linear_3.f90	3-neuron linear network employing the Adam gradient descent optimisation algorithm
linear_3_nadam.f90	3-neuron linear network employing the Nadam gradient descent optimisation algorithm
linear_8_att.f90	8-neuron linear network employing the Adam gradient descent optimisation algorithm and element-wise dot-product attention
linear_8_mh_att.f90	8-neuron linear network employing the Adam gradient descent optimisation algorithm and multi-head attention
sompr_5.f90	Second order multivariate polynomial regression employing the Adam gradient descent optimisation algorithm
gelu_att.f90	GELU activation function, Adam gradient descent optimisation algorithm, element-wise dot-product attention
mish_att.f90	Mish activation function, Adam gradient descent optimisation algorithm, element-wise dot-product attention
smish_att.f90	Smish activation function, Adam gradient descent optimisation algorithm, element-wise dot-product attention
gms_att.f90	GELU + Mish + Smish activation functions, Adam gradient descent optimisation algorithm, element-wise dot-product attention

fortran_ai's Issues

Used name constants for neural network size

In your code gelu_att_8_8_8.f90 you hard code the network size 8 in many places. It would make the code more to flexible to define something like

integer, parameter :: net_size = 8

and use net_size instead of 8 in the code. The same suggestion applies to the other codes.