Reinterpretation of axes. Right now, our axes are cost and cost + cal*error^2. What does it mean? Try to think of it as a thought experiment so when we present it we captivate the audience about how this is important or relevant. Or, another plot: rewards vs effort costs, utility is rewards minus cost
Figure out what we want to highlight for our presentation. Higher-level conclusions + overview of methods. The analogies.
After Tuesday:
Information complexity in hierarchical model. It should be cheaper to keep around abstract information (such as category) instead of precise information (such as the weight of a specific coin). The hierarchies with more categories are more complex. Knowing the exact weight informs probability of category because of conditioning on child provides information about parent.
What if you noisify your category belief, how does that effect your ability to generalize to other coins.
Before Tuesday:
After Tuesday: