Day6: The Neuron

The sklearn.fit() Trap

Here is the version of this lesson you will find in 90% of beginner ML tutorials:

Three lines. It works. And it teaches you almost nothing about what a neuron
actually does.

What sklearn hides: the weighted sum computation, the bias addition, the
threshold comparison, the weight update rule, and the convergence condition.
Every one of those steps is a design decision with mathematical consequences.
When your model misbehaves in production — and it will — you will have no
mental model for why, because the library never let you see the mechanics.

This lesson costs you three more files and about 80 lines of NumPy. What you
gain is a complete, transparent mental model of the most fundamental unit in
all of deep learning.