Package Features

Customizability

Customizability

All of the neurons, layers, and components will/ can have customizable functions that can be used when the network is built. More specifically, the Activation, Delta, Loss functions, etc... will be customizable.

Fault-Tolerance

Fault-Tolerance

Using GenServer, all entities in the neural net are treated as processes with a state and are managed such that if they fail, they are resumed.

Few Maths

Few Maths

Leveraging the benefits of the Erlang BEAM engine, no matrix math is used to perform calculations for the neural net. Instead, a more direct approach is used that is modeled off a visual diagram of a neural net.

Open Source

Open Source

The project is and will continue to remain open source. Requests for features will be considered and potentially implemented by/for the user community.

Roadmap to 1.0

What can you expect from the 1.0 release?

With the 1.0 release, you can expect a functioning library that has full support for the most popular functions that are used for neural nets. We expect to get some community involvement to help us tailor the library to the people that are going to use it! The system is going to be build with a very open architecture in mind to simplify changes and feature additions. It will remain compatible with the newest versions of Elixir and the OTP.

See the published package on Hex.pm

Versions

References

  • "GenServer – Elixir v1.7.1", Hexdocs.pm, 2018. [Online]. Available: https://hexdocs.pm/elixir/GenServer.html. [Accessed: 01- Aug- 2018].
  • B. Widrow, D. Rumelhart and M. Lehr, "Neural networks: applications in industry, business and science," Communications of the ACM, vol. 37, (3), pp. 93-105, 1994. . DOI: 10.1145/175247.175257.
  • "Platform features", https://elixir-lang.org/, 2018. [Online]. Available: https://elixir-lang.org. [Accessed: 01- Aug- 2018].
  • "Phoenix", Phoenix Framework, 2018. [Online]. Available: https://phoenixframework.org/. [Accessed: 01- Aug- 2018].
  • "Erlang -- What is Erlang", Erlang.org, 2018. [Online]. Available: http://erlang.org/faq/introduction.html. [Accessed: 01- Aug- 2018].
  • F. Hunleth, "Craft and deploy bulletproof embedded software in Elixir — Nerves Project", Nerves-project.org, 2018. [Online]. Available: https://nerves-project.org. [Accessed: 01- Aug- 2018].
  • "Electron", Electronjs.org, 2018. [Online]. Available: https://electronjs.org. [Accessed: 01- Aug- 2018].
  • "Discord", Discord, 2018. [Online]. Available: https://discordapp.com. [Accessed: 01- Aug- 2018].
  • "WhatsApp", WhatsApp.com, 2018. [Online]. Available: https://www.whatsapp.com. [Accessed: 01- Aug- 2018].