*   `CategoryCrossing`: create new categorical features representing
        co-occurrences of previous categorical feature values
    *   `Hashing`: the hashing trick, for large-vocabulary categorical features
    *   `Discretization`: turn continuous numerical features into categorical
        features by binning their values
*   Improved **image preprocessing** layers: `CenterCrop`, `Rescaling`

Since these new workflows don’t require any new cluster features, they’re available for experimentation with your existing clusters via `kubectl alpha debug`. We’d love to hear your feedback on debugging with `kubectl`. Reach us by opening an issue, visiting [#sig-cli](https://kubernetes.slack.com/messages/sig-cli) or commenting on enhancement [#1441](https://features.k8s.io/1441).

### Structured logging