Sophelen Research

A few years ago, I was working on using basic machine learning for aquaponics, but I quickly gave up due to its weaknesses. Basic machine learning does not integrate the geometry of the data, which is why geometric-based ML is popular.

I’ve been trying to find ways to work directly with the aquaponics manifold. The manifold hypothesis tells us that no matter how many variables we decide to use to track it, its effective dimension is low. As aquaponics follows the laws of nature, the manifold itself should be smooth, unlike the supply chain manifold.

After listening to many amazing podcasts from Mr. Scott Luton from Supply Chain Now, I’ve decided to see if I can work the supply chain manifold directly via LETKF, a powerful tool from geoscientists.