![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
I was wondering if Other People who have not had some silly amount of training in math/science perceive figuring out a recipe this way. Because I cannot help but think about this every time I cook and of course my bf has the same training so instead of like, discussing if this is normal we just fall into discussions of why he thinks cooking is more like active learning (iterative supervised learning, aka machine learning nonsense) than steepest descent, which is what /I/ think of it as.
OK so: steepest descent. Imagine salt is an axis on which you can move (by adjusting the amount used). The curve created by how tasty the food is at each level of salt is the energy landscape. Let's call it the tasty landscape. The steepest descent algorithm would say, at this point on the tasty landscape, which direction (more or less salt) would be tastier (the gradient). Therefore the next time we make the recipe we will update in that direction (the direction of steepest descent). Now extrapolate to all the various ingredients / methods of preparation as the axes, and you get a full tasty landscape upon which you can use the steepest descent algorithm.
(It's a descent bc usually the lower energy / more stable form is desired. I guess if you think of higher numbers as tastier, it would be a steepest ascent.)
For example, in my chili recipe, there's lots of spices etc, so I'm varying the quantities and ... existence of spices and enacting the steepest descent algorithm when I go: oh, the direction to make this tastier is more bay leaf and less star anise. But this tasty landscape is multidimensional, so it's a difficult problem for me to assess the gradient of!
Anyway, I'm a nerd, is anyone the same kind of nerd as me?
OK so: steepest descent. Imagine salt is an axis on which you can move (by adjusting the amount used). The curve created by how tasty the food is at each level of salt is the energy landscape. Let's call it the tasty landscape. The steepest descent algorithm would say, at this point on the tasty landscape, which direction (more or less salt) would be tastier (the gradient). Therefore the next time we make the recipe we will update in that direction (the direction of steepest descent). Now extrapolate to all the various ingredients / methods of preparation as the axes, and you get a full tasty landscape upon which you can use the steepest descent algorithm.
(It's a descent bc usually the lower energy / more stable form is desired. I guess if you think of higher numbers as tastier, it would be a steepest ascent.)
For example, in my chili recipe, there's lots of spices etc, so I'm varying the quantities and ... existence of spices and enacting the steepest descent algorithm when I go: oh, the direction to make this tastier is more bay leaf and less star anise. But this tasty landscape is multidimensional, so it's a difficult problem for me to assess the gradient of!
Anyway, I'm a nerd, is anyone the same kind of nerd as me?
