Machine Learned “Initial Temp” Calculator for Beer Cooler Sous-Vide
The summer sous-vide machine learning project has paid off. I now have enough data to estimate the desired initial starting temperature for the water, given the amount of water in the cooler, a history of the cooler’s temperature logs, the starting mass of the food, the desired stasis temperature and the desired cooking time.
The machine learning algorithms estimate the hidden temperature loss function of a beer cooler (sous-vide cooking vessel) well enough so that “correct” cooking can be approximated without a heating element. There are downsides to version 0.1 of this technique — it only works well in a small temperature range (100 to 140 degrees F) and it hasn’t been tested on a lot of different vessels.
Once I have some more success with the calculator on predicting out of sample (in real cooking situations), I will publish the calculator on the web for others to use.
