Neural Network Bulder

This application allows new neural network developers to play around with network architecture and to train/run neural network. Built from scratch, this beginning neural network utilizes a sigmoid activator function and stochastic gradient descent. The UI of the neural network allows users to customzie the architecture of a neural network without ever touching the code behind a neural work or attemping to omptimize layer usage and other critical features of a neural network that affect the effectiveness of a neural network.

bandicam.2021-05-01.23-30-14-762_Trim.mp4

Dependencies

• clang
• cmake
• cinder

Windows

Visual Studio

• Visual Studio 2017+

Quick Start

• Download Cinder (latest version)
• Create a new project within the Cinder directory

Builder Controls

Keyboard

Key	Action
Right arrow	Adds a layer to the end
Left arrow	Removes a layer from the end
Up arrow	Adds a neuron to the last layer
Down arrow	Removes a neuron from the last layer

File Drops

The file drop event is meant to receive a text file with training or predict data. You must drop Training Data in before you can use predict features!

• Key represents the text at the top of the text file

Key	Action
TRAIN	Trains the neural network with the file data
PREDICT	Predicts the data and displays the neural network output layer values in the outpout neruon

Train file structure

TRAIN
EXPECTED_VALUE_1 VALUE_1 VALUE_2 ...
EXPECTED VALUE_2 VALUE_1 VALUE_2 ...
.
.
.

Train file structure

PREDICT
VALUE_1 VALUE_2 ...

This application was developed by Amit Sawhney. Last updated May 2021.