Artificial Intelligence By Example

Shuffling as an alternative to random sampling

K-means clustering is an unsupervised training algorithm. As such, it trains unlabeled data. One random computation does not consume a high level of machine resources, but several random selections in a row can.

Shuffling can reduce machine consumption costs.

Proper shuffling of the data before starting training, just like shuffling cards before a Poker game, will avoid repetitive and random mini-batch computations. In this model, the loading data phase and training phase do not change. However, instead of one or several random choices for dataset1, the mini-batch dataset, we shuffle the complete dataset once before starting the training. The following code shows how to shuffle datasets.

sn=4999 ...