Home

NNN - Neural Network Notions - Documentation

• Installation

  • Standard DLL Download

  • NuGet Package

  • Unity Package

• Compatibility

  • Language, Hardware, and OS Requirements

  • .NET Compatibility

  • Unity Compatibility

• Getting Started

  • Saving / Loading a Model

  • Training a Standard Model

  • Training a DQN Agent

  • Creating a Custom DQN Environment

• C# API

  • Trainers

  • Optimizers

  • Cost Functions

  • Models

  • Layers

  • Activation Functions

  • Tensors

  • Environments

  • Buffers / Records

  • Save System

• Core Design Principles

  • Development Goals / Aims

  • Design Philosophy

• Mathematical Definitions

  • Element-Wise Tensor Operations

  • Complex Tensor Operations

  • Activation Functions

  • Cost Functions

  • Optimizer Functions

  • Miscellaneous Functions

• Repository Organization

• C# Implementation

  • Activations

  • Autodiff

  • Buffers

  • Compiler Attributes

  • Costs

  • DQN Environments

  • Episodes

  • Interop

  • Models

  • Layers

  • Optimizers

  • Trainers

  • Utilities

  • Data Loaders

  • Save System

• C# / C++ Barrier

  • Interop

  • Memory Management / Preventing Leaks

• C++ Implementation

  • Tensors

  • Autodiff Engine

  • Other Neural Network Functions

  • SIMD Vectorization

  • Parallelization

  • Compiler Settings

• File Format (.nnn)

  • General Formatting Notes

  • File Header Format

  • Model Header Format

  • Layer Header Format

  • Layer Data Format

  • Activation Function Data Format

  • Tensor Format

  • Primitive Data Type Formats

  • Layer IDs

  • Activation IDs