Three pronged attack for the next 4 weeks:
Sometime next week need to dedicate some time to make things installable.
On a cadence of once every 3-5 days produce 2 ideas for new features based off of learnings.
Notes on triggering:
1. run match from frontend, no stoers are changed
2. on backend, match creates new column, and returns storemodification
3. on frontend, store modification is applied to the store (ideally this should stop)
4. (without trigger) on the frontend, store triggered is called
5. on the backend the store is updated, and the store_modification is sent to the front end
6. on the frontend, the value in the store modification is the same as teh value of the store – so no trigger is sent
Interfaces we’d like to implement:
dp["label"] = np.ones(len(dp)) * -1
dp["sunglasses1"] = match("...")
dp["label"][dp["sunglasses1"] > 0.2] = 1
dp["label"][dp["sunglasses1"] < -0.5] = 0
dp["sunglasses2"] = match("...")
dp["label"][dp["sunglasses2"] > 0.1] = 1
dp["sunglasses3"] = match("...")
dp.filter("unlabeled")["label"][dp["sunglasses3"] > 0.1] = 1
# manul item labeling
dp[12] = 1
dp[13] = -1
dp[15-19] = 1
# interaction
dp.sort("labeled") # all the negative ones go to the bottom
dp.filter("labeled_positive")
dp.filter("labeled_negative")
dp.filter("unlabeled")