You can run this notebook in a live session Binder or view it on Github.

Train Models on Large Datasets

Most estimators in scikit-learn are designed to work with NumPy arrays or scipy sparse matricies. These data structures must fit in the RAM on a single machine.

Estimators implemented in Dask-ML work well with Dask Arrays and DataFrames. This can be much larger than a single machine’s RAM. They can be distributed in memory on a cluster of machines.

In [1]:
%matplotlib inline
In [2]:
from dask.distributed import Client

# Scale up: connect to your own cluster with more resources
# see
client = Client(processes=False, threads_per_worker=4,
                n_workers=1, memory_limit='2GB')



  • Workers: 1
  • Cores: 4
  • Memory: 2.00 GB
In [3]:
import dask_ml.datasets
import dask_ml.cluster
import matplotlib.pyplot as plt

In this example, we’ll use dask_ml.datasets.make_blobs to generate some random dask arrays.

In [4]:
# Scale up: increase n_samples or n_features
X, y = dask_ml.datasets.make_blobs(n_samples=1000000,
X = X.persist()
dask.array<concatenate, shape=(1000000, 2), dtype=float64, chunksize=(100000, 2)>

We’ll use the k-means implemented in Dask-ML to cluster the points. It uses the k-means|| (read: “k-means parallel”) initialization algorithm, which scales better than k-means++. All of the computation, both during and after initialization, can be done in parallel.

In [5]:
km = dask_ml.cluster.KMeans(n_clusters=3, init_max_iter=2, oversampling_factor=10)
KMeans(algorithm='full', copy_x=True, init='k-means||', init_max_iter=2,
    max_iter=300, n_clusters=3, n_jobs=1, oversampling_factor=10,
    precompute_distances='auto', random_state=None, tol=0.0001)

We’ll plot a sample of points, colored by the cluster each falls into.

In [6]:
fig, ax = plt.subplots()
ax.scatter(X[::1000, 0], X[::1000, 1], marker='.', c=km.labels_[::1000],
           cmap='viridis', alpha=0.25);

For all the estimators implemented in Dask-ML, see the API documentation.