pyTAGI review: How does the backward(.) work

[@guangjunxu5325]

👋👋👋👋👋👋👋👋👋👋👋👋👋👋👋👋👋👋👋👋
Thank you sincerely for the detailed explanation. I really like how you clarified the Kalman gain calculation based on the GMA (Gaussian Multiplicative Approximation) and the mean-field assumption. It was incredibly clear and insightful.
And really learned something new-it's interesting to note that the output layer should be calculated independently rather than updated together with the hidden layers. I hadn't considered this distinction before, but now it makes a lot more sense.