BrabeNetz - Pre-built x64 Release library (MSVC2017 and g++)

Pre-built binary library (.lib/.a) with MSVC2017 and g++, on Release for x64

New Features:

• ➕ New modern C++ proxy class (wrapper) for better neural network interaction
• ➕ Preprocessor directives changed to runtime structs/properties
• 🔥 Faster backpropagation
• 🔥 Faster initialization allocation
• 🖌️ Code cleanup

Contents:

• Linux (g++ c++17):
  • headers/*.h: All BrabeNetz headers for symbols
  • lib/BrabeNetz.a: C++ static class library
  • trainer/BrabeNetzConsole: Console trainer application for MNIST Image recognition
  • trainer/state.nn: Pre-trained neural network for image recognition on {784, 500, 100, 10}
• Windows (MSVC2017):
  • headers/*.h: All BrabeNetz headers for symbols
  • lib/BrabeNetz.lib: C++ static class library
  • trainer/BrabeNetzConsole.exe: Console trainer application for MNIST Image recognition
  • trainer/state.nn: Pre-trained neural network for image recognition on {784, 500, 100, 10}

Trainer:

The trainer is configured to run MNIST handwritten character recognition, multithreaded, and output all Images + Status updates to the console. To efficiently train the network, you might want to disable output and recompile yourself.

Be aware: To run the training app, you have to download the MNIST training sets and put them in the same folder as BrabeNetzConsole app. (images, labels)

BrabeNetz - Pre-built x64 Release library (MSVC2017 and g++)

Pre-built binary library (.lib/.a) with MSVC2017 and g++, on Release for x64

New Features:

• ➕ Compile for Linux! (CMake/CLion)
• ➕ Preprocessor directives changed to runtime structs/properties
• 🔥 Faster backpropagation
• 🔥 Faster initialization allocation
• 🖌️ Code cleanup

Contents:

• Linux (g++ c++17):
  • headers/*.h: All BrabeNetz headers for symbols
  • lib/BrabeNetz.a: C++ static class library
  • trainer/BrabeNetzConsole: Console trainer application for MNIST Image recognition
  • trainer/state.nn: Pre-trained neural network for image recognition on {784, 500, 100, 10}
• Windows (MSVC2017):
  • headers/*.h: All BrabeNetz headers for symbols
  • lib/BrabeNetz.lib: C++ static class library
  • trainer/BrabeNetzConsole.exe: Console trainer application for MNIST Image recognition
  • trainer/state.nn: Pre-trained neural network for image recognition on {784, 500, 100, 10}

Trainer:

The trainer is configured to run MNIST handwritten character recognition, multithreaded, and output all Images + Status updates to the console. To efficiently train the network, you might want to disable output and recompile yourself.

Be aware: To run the training app, you have to download the MNIST training sets and put them in the same folder as BrabeNetzConsole app. (images, labels)

BrabeNetz - Pre-built x64 Release library (MSVC2017 and g++)

Pre-built binary library (.lib/.a) with MSVC2017 and g++, on Release for x64

New Features:

• ➕ Compile for Linux! (CMake/CLion)
• 🖌️ Code cleanup

Contents:

• Linux (g++ c++17):
  • headers/*.h: All BrabeNetz headers for symbols
  • lib/BrabeNetz.a: C++ static class library
  • trainer/BrabeNetzConsole: Console trainer application for MNIST Image recognition
  • trainer/state.nn: Pre-trained neural network for image recognition on {784, 500, 100, 10}
• Windows (MSVC2017):
  • headers/*.h: All BrabeNetz headers for symbols
  • lib/BrabeNetz.lib: C++ static class library
  • trainer/BrabeNetzConsole.exe: Console trainer application for MNIST Image recognition
  • trainer/state.nn: Pre-trained neural network for image recognition on {784, 500, 100, 10}

BrabeNetz - Pre-built x64 Release library (MSVC2017, `.lib`)

Pre-built binary library (.lib) with MSVC2017, on Release for x64

New Features:

• 🔥 Faster backpropagation
• 🔥 Faster initialization allocation
• ➕ network::build_topology() function; updating and returning a network_topology
• 🖌️ Code cleanup
• ➕ Improve multithreading worth-it-checker for spawning threads

Contents:

• BrabeNetz.lib: C++ class library
• *.h: All BrabeNetz headers for symbols

XOR Pre-Built Network and Library

A { 2,5,11,30,80,100,80,30,11,5,1 }, training a XOR 4.000.000 times.

Contents:

• Pre-Built .exe (Console Application) with network running above specs
• Pre-Built .lib (Static library) (newest build)
• Header (.h) files

XOR Pre-trained Network

A compiled network pre-trained to solve XOR (state.nn)