This is a flexible low level library that allows for development of innovative update rules in the context of closed-loop deep learning. It can be used with the conventional back-propagation algorithm or the newly developed 'local propagation of global (closed-loop) error' algorithm. This repository is intended for use as an external library to any learning applications.

you can find descriptions of all functions in the doxygen output file refman.pdf

CLDL uses cmake. just enter the CLDL directory from the root:

• cd CLDL

and type:

• mkdir build && cd build
• cmake ..
• make

record the path to both the generated library file (libCLDL.a) and of the include directory for external use in other projects.

A Unit test is included in the tests directory that shows how the library is used for learning with back-propagation. The executable tests will be generated automatically when building CLDL. Run the test by doing:

• cd tests
• ./tests

GNU GENERAL PUBLIC LICENSE

Version 3, 29 June 2007

(C) 2018,2019,2020 Sama Darya <[email protected]>