QuantNeuroChaOSys.js Logo

QuantNeuroChaOSys.js - Еxample

Automatic Training and Solving a Problem with a Neural Network

In this example, we train a neural network on numerical data and use it to predict the next element in a sequence.

Results:

Prediction Result for Training Data (1-10):


    

Prediction Result for Test Data (11-15):


    

Generated Data for Training:


    

Test Data:


    

Source Code of Example:


    

What do these results mean?

Training Results: The neuro-quantum network is trained on numerical data to learn how to predict the next element in the sequence. Training doesn't always result in an exact match with the target values, but it is important that the neural network can capture the trend or pattern (e.g., predictions for numbers 1-10, such as 10.25, 18.47, etc.). This is normal behavior for most neural networks as they focus on general trends rather than exact data reproduction.

Test Data: When the neuro-quantum network is tested on data it hasn't seen during training (e.g., for numbers 11-15), it continues making predictions based on the trend it has learned. While the predictions may deviate slightly from the expected values (e.g., for 11 — 92.47, for 12 — 100.70), they still follow the overall pattern, indicating good generalization of the neural network.

Prediction Errors: Errors in predictions can arise for various reasons, including randomness in the training data or the complexity of the task. However, if the predictions follow the trend and the error is small, this is considered a successful training result.

Why This is Normal: Neural networks don't always exactly reproduce the original data, especially if it is subject to random changes or noise. In real-world applications, it is important for the model to make reasonable predictions for new data, rather than simply memorizing the old ones. Predictions that follow the trend (e.g., increasing by the same amount) usually indicate good generalization capability of the model.