Revision: Figure 4 -- New Figure 5

[jaybee84]

Please take a look at this figure from the point of view of an audience less familiar with ML. What are some of the points that are confusing?

[allaway]

I think that there are a few things we could change to make this figure more understandable:

• Explain what the colors and shapes mean. Maybe, additionally, we could remove the "dataset" concept and just leave the "identity" concept
• Rearrange the figure somewhat. The training data is input to the untrained model, and the rare disease data is input to the trained model. As organized, it kind of looks like the prior data becomes the trained VAE/PLIER, which outputs representations, which then somehow become the test data? We should make it clear what the input and output to the trained models are.
• B) heterogeneous is missing an E :)

[dvenprasad]

Current figure for reference:

[jaclyn-taroni]

Noting here that I dropped a sketch of some thoughts in the InVision Freehand to the right of the transfer learning sketches, figures, etc.

[jaybee84]

@dvenprasad can you please add the new figure to the repo?

[dvenprasad]

I think we're still iterating on this. But this is what we have so far.