euler's method
1. similarity with gradient descent
- In an undergrad ODE class, you solve initial value problems of the form \(y'(t) = f(t, y(t))\) and then to solve this numerically using Euler's method you do \(y_{n+1} = y_n + hy'_n\).
- With gradient descent you have a problem of the form \(y'(t) = - \nabla F(y(t))\). Gradient descent solves the problem of finding a "gradient flow"
- taken mostly from reddit comment here