AI crossword

2. The type of matrix whose eigenvectors define the principal components in PCA
6. In PCA, this quantity gives the variance captured along a principal component
7. S-shaped function that squashes a linear score into a probability between 0 and 1
8. In reinforcement learning, the signal an agent tries to maximise
9. The direction of steepest descent/ascent that an optimisation algorithm steps against
10. Measures the systematic error or the distance between a model's average predicted values and the true/actual values (ground truth)

1. The recorded history of where a piece of data came from and how it was produced
3. Training many trees on bootstrap samples and averaging their predictions
4. Of all the instances a classifier flags as positive, the fraction that are genuinely positive
5. When a model memorises training noise instead of the underlying pattern, shown by a large train–validation error gap