Neural network made with C++ (info below):

This is a neural network for binary sentiment analysis made in C++. The data (Sentiment140 dataset) was cleaned using Python/Jupyter notebook. All of the paths to files have been changed/omitted because I like keeping my desktop contents secret.

First Attempt:

• 1 hidden layer.
• 50 hidden nodes.
• ReLU for hidden node activation.
• Sigmoid for output normalization.
• Backpropagation is not implemented correctly.
• Atypical node connection.
• 1 file.

Second Attempt:

• 2 layers.
• 50 nodes in hidden layer 1.
• 25 nodes in hidden layer 2.
• ReLU for all hidden node activation.
• Sigmoid for output normalization.
• Backpropagation implemented somewhat correctly.
• Atypical node connection.
• 1 file.

Third Attempt:

• 2 layers.
• 25 nodes in hidden layer 1.
• 25 nodes in hidden layer 2.
• ReLU used for all hidden node activation.
• Sigmoid for output normalization.
• Backpropagation implemented correctly.
• ADAM used for backpropagation.
• Xavier initialization used.
• Multiple files.