ml

3. Algorithm used to update weights during training
7. The term for adjusting weight values using gradients
10. When market or consumer behaviour changes, causing model accuracy to drop
13. The final dataset used to evaluate general performance
14. The subset of data used for hyperparameter tuning and early stopping
16. One complete cycle through the training dataset
18. Neural network element that receives weighted inputs
19. A problem where a model fits too well to training data but performs poorly on new data
21. The process of dividing data into training, validation, and testing subsets
23. A computational model inspired by the human brain used for prediction and learning
24. When a model is too simple and fails to capture business patterns
25. Number of samples processed before weights are updated

1. When a trained model memorizes the training data too closely and performs poorly on the testing dataset
2. Error resulting from overly simplistic models
4. Function measuring how far predictions deviate from actuals
5. The phenomenon where model performance worsens due to changes in market or business conditions
6. Evaluation method that mitigates dependence on one specific train–test split
8. Provides nonlinearity inside a neural network
9. A matrix of pixel values fed into a neural model
11. Input and output example pair used during training
12. A plot comparing training and validation error across epochs
15. Opposite of overfitting
17. Error caused by overly complex models
20. Popular activation function used in modern neural networks
22. Direction of steepest descent during optimization