search

josephjaspers / blackcat_tensors

Matrix-Vector Library Designed for Neural Network Construction. cuda (gpu) support, openmp (multithreaded cpu) support, partial support of BLAS, expression template based implementation PTX code generation identical to hand written kernels, and support for auto-differentiation
12 stars 4 forks source link

Documentation for Neural Network module #24

Closed josephjaspers closed 4 years ago

josephjaspers commented 4 years ago

Currently lacking docs on neural network module. Need to add.

josephjaspers commented 4 years ago

https://josephjaspers.github.io/BlackCat_Tensors_Doxygen/html/namespaceBC_1_1nn.html

josephjaspers commented 4 years ago

Neural Network documentation will use doxygen.

The Tensor library will be a mixture of doxygen and self-made docs.

Reasoning: The tensor library's feature set is much larger than the neural-network module. Manual documentation enables us to highlight the more important aspects of the Tensor class. The neural-network libraries feature-set is much smaller (even though it depends on the tensor-library). Additionally the features used in the neural-network library would be fairly clear to anyone with a background in NNs and data-science.

The tensor library however would require user's to have a moderate understanding of both openmp and CUDA which makes it inherently unfriendly to non-expert users.

josephjaspers commented 4 years ago

Moving NNs to different project