AI Crossword

2. A trainable parameter that shifts a neuron’s activation independent of inputs.
4. One full pass through the entire training dataset during learning.
7. A function that measures how far a model’s predictions are from targets.
8. A computing model made of interconnected layers that mimics neuron connections in the brain.
11. The vector of partial derivatives indicating how to change parameters to reduce loss.
14. The field where algorithms learn patterns from data without explicit programming.
16. When a model learns training noise and performs poorly on new data.
18. A nonlinear function applied to a neuron’s output to introduce complexity.
19. Techniques that reduce overfitting by constraining model complexity.

1. The process of breaking text into pieces (tokens) for model input.
2. The method for computing gradients to update neural network weights.
3. A learning paradigm that trains on labeled input–output pairs.
5. A learning paradigm that finds structure in unlabeled data.
6. A step-by-step procedure for solving a problem or performing a task.
9. A dense vector that represents discrete items (words, tokens) in continuous space.
10. A model component that generates output from a latent representation.
12. A model component that converts raw input into a latent representation.
13. A neural architecture that uses self-attention to handle sequences.
15. A trainable parameter that scales an input in a neural network.
17. A structured collection of examples used to train or evaluate models.