Seven Segment Neural Network

A neural network trained to recognise seven segment 'digital clock' digits.

Neural Network Diagram
Neural Network diagram showing 2 hidden layers

Network Architecture

The model is a simple neural network that maps binary input (if the LED is on or off) patterns from seven-segment displays to digits. Data passes sequentially through two hidden layers before producing an output.

ReLU Activation Function
Graph showing ReLU activation function

Activation Function

The hidden layers use the ReLU (Rectified Linear Unit) activation function. ReLU introduces non-linearity into the network, which lets it learn more complex patterns.
In this case, it helps the network capture subtle relationships between which segments are on or off and the digit they represent.

Loss curve of NN training
Loss curve showing how the network becomes more accurate during training

Training the Network

The network was trained on all 10 digits, encoded as binary patterns for each seven-segment input. Supervised learning was used, with the model optimising a loss function to reduce prediction errors.
The loss curve shows this process, highlighting how the error decreases over epochs (how many times the model was run) and how the network gradually becomes more accurate.

Accuracy of NN predicting digits
Accuracy of the neural network predicting each digit

Model Performance

After training, the network achieves high accuracy in recognising digits from the input patterns. This can be visualised through the accuracy graph, highlighting how well the network accurately guesses each digit.

Learnings and Insights

Encoding input effectively (binary representation of segments).
Understanding hidden layer transformations and non-linear activations.
Monitoring training progress using loss and accuracy curves.

This NN shows a small example of supervised learning, pattern recognition and the potential it has.