Across
- 1. Final prediction of a neural network
- 2. Activation function that outputs zero or positive values
- 4. added constant term in neural unit
- 8. Algorithm to update weights via error propagation
- 10. One complete pass through the training data
- 12. Data fed into a neural network
- 13. function that introduces non-linearity
Down
- 1. When model memorizes instead of generalizing
- 3. Measures error between prediction and target
- 5. Vector of partial derivatives used for optimization
- 6. Regularization technique dropping random neurons
- 7. Intermediate layer between input and output
- 9. Method like SGD or Adam to minimize loss
- 11. Activation function that outputs between 0 and 1