We currently support the following strategies for reduction from forecasting to regression:

• direct
• recursive
• dirrec (see #226)

Some models can directly predict multiple outputs, see e.g. LinearRegression:

import numpy as np
from sklearn.linear_model import LinearRegression

y = np.random.normal(size=(10, 3))
X = np.random.normal(size=(10, 5))

estimator = LinearRegr

According to Section 3.4 of the aamp paper, the non-normalized Euclidean distance (i.e., L2-norm) can be extended to p-norm which includes Manhattan distance (p=1). We can follow the numpy.linalg.norm syntax and add an order parameter (ord is a reserved keyword) that accepts a positive integer for computing the p-norm.

Hi, Thanks for the awesome library!

So I am running a Kmeans on lots of different datasets, which all have roughly four shapes, so I initialize with those shapes and it works well, except for just a few times. There are a few datasets that look different enough that I end up with empty clusters and the algorithm just hangs ("Resumed because of empty cluster" again and again).

I conceptually

There are way to many UserWarnings coming out of the testing suite (causing at least one of the travis jobs to terminate). Ideally the tests should avoid running into those issues except in the case when their raising is tested on a context manager.

I suppose this issue can be a nice starting point for someone new to the project and wanting to become familiar with the code base before e.g. GSoC