Скачать или смотреть BigML - Streaming Histogram (wine alcohol content vs quality)

This video shows a streaming histogram in action. The histogram uses a fixed amount of memory (16 bins in this example) and dynamically allocates bin positions and widths during one pass over the data.

The histogram is extended so that it can capture information about a second feature. This lets us see potential correlations between variables.

The plot shows the distribution of wine alcohol content for a variety of white wines (the red line) versus the average wine quality rating (the blue line). Each frame is rendered after the histogram is updated with data about 5 more varieties. In the last frame, the histogram has observed 2500 kinds of wine.

The data is also streaming into the histogram in sorted order. This can be tricky for some streaming histogram techniques, but this algorithm handles it successfully.

Thanks to Y. Ben-Haim and E. Tom-Tov for publishing the algorithm:
http://jmlr.csail.mit.edu/papers/v11/...

Thanks to S. Tyree, et. al. for the histogram extension:
http://research.engineering.wustl.edu...

Thanks to the UCI Machine Learning Repository for the data:
http://archive.ics.uci.edu/ml/dataset...