Gabe

set it up so the static medical numbers are read from the config file and passed to the synthetic data generating/model generating code.

test.py does not parse arguments/config file so is unusable currently.

While training, collect the number of false signs and symptoms. (Can we even do this?) This would be helpful in adding a negative component to missing scores in the differential...

without storing the function variables on the graph, it becomes hard to store the edges as there is no good way to serialize the functions that are currently stored. Suggest...

Build an ability to recommend treatments that have positive effects and no negative effect on the potential diagnoses.

Build the ability to recommend tests based off of their ability to differentiate between signs/symptoms.

Calculate tau2 or such on the distribution and only keep edge counts above it. If it fails, use the 95th percentile or such.

Subtract an amount from a diagnosis's score for every sign/symptom in the record that is missing from the diagnosis. how much to subtract is the question. At fault of one...

When displaying scores, values should be sorted by their percent difference. so if the first item gets a score of .4 and the second a score of .3, the scores...