doll

D Optimization Library for Learning - A doll you can play with, dress it up and imitate some real-life situations, as if it was alive.

doll is a collection of several gradient-free and gradient-dependent optimization algorithms. The primary goal of this library was to test the usability of the D language for such mathematical purposes.

Documentation

For API reference you can generate the library documentation from the source code using one of the D documentation generators (ddoc, ddox, adrdoc etc.). For example run:

dub build -b ddox

Testing

To test the implemented optimization functions you can write your own test project or use the example codes from the documentation. These are also unit test sections therefore you can easily test all of them with the following command.

dub test

Or if you are using ldc:

dub test --compiler=ldc2

TODO list

• Implement ADAM optimizer (gradient dependent)
• Optional parallelism in certain calculations.
• OTHER: Use D's recurrence and sequence methods (and ranges in general) to express mathematical formulas

License

The library is distributed under the terms and conditions of GNU LGPL v3.0. Copyright © 2019, 0l-l0