Supporting material for my OSDC London'16 talk on Dataset Shift in Machine Learning
Interactive plot to illustrate the phenomenon of covariate shift. That is, p(x,y) differs from training to testing phase. The interactive plot allows you to specify p(x) via a probability tabels while p(y|x) remains fixed. You can see that underspecified discriminative models are not immune to covariate shift.
Use the bokeh serve
command to run the interactive plot:
bokeh serve checkerboard/bokeh_checkerboard.py
at your command prompt. Then navigate to the URL
http://localhost:5006/bokeh_checkerboard
in your browser .