Migrate all Python code from old-fashioned format() functions, formatting % operators and simple concatenations (+) to modern f-strings (brief guide). They are known to be the fastest approach and also increase code readability.

![image](https://user-images.githubusercontent.com/25141164/112898582-a

Refer to doc2vec.py, infer_vector function seems to be using epochs for the number of iterations and steps is not in used.

However, in the similarity_unseen_docs function, steps is used when calling the infer_vector function.

Problem: the approximate method can still be slow for many trees
catboost version: master
Operating System: ubuntu 18.04
CPU: i9
GPU: RTX2080

Would be good to be able to specify how many trees to use for shapley. The model.predict and prediction_type versions allow this. lgbm/xgb allow this.

The official instructions say to use joblib for pickling PyOD models.

This fails for AutoEncoders, or any other TensorFlow-backed model as far as I can tell. The error is:

>>> dump(model, 'model.joblib')
...
TypeError: can't pickle _thread.RLock objects

Note that it's not sufficient to save the underlying Keras S

LET page = DOCUMENT("YOUR_URL", {
	driver: "cdp",
	ignore: {
        resources: [
            {
                url: "*",
                type: "image"
            }
        ]
    }
})

The classifiers have many and varied parameters. Most could be improved with

1. Clearer description of the parameter meaning
2. Action/exceptions thrown when the classifier is parameterised in a way that will make it crash

For example, see this issue for the RISE classifier.
alan-turing-institute/sktime#895

some checks can be done when the classifier is constr

• What's your use case?