Machine learning --Mini Batch K-Means algorithm

　　Mini Batch K-Means algorithm is a K-Means algorithm optimized variant, using small subset of data (each set of training data used in the training algorithm is randomly selected subset of the data) reduce the computing time, while trying to optimize the objective function; Mini Batch K-Means algorithm can reduce the convergence time K- Means algorithm, and the results produced results only slightly worse than the standard K-Means algorithm

Algorithm steps as follows:
　　● extracted first part of the data set, using K-Means algorithm to construct the K clusters model points
　　● continue to take training | training data set portion of the sample data set, and adds it to the model, is assigned to nearest cluster center point
　　● updated cluster center point value (only updated with each extracted partial data sets)
　　● the second and third loop iteration step operation until the number of iterations reaches the center point or stability, cease to operate

For example, 10,000 samples each take 1000 samples, do a K-means, without replacement.

　　Then take 1000 ......