# How Neural Networks work
* Every nueron has an activation function and synapses have weights
* To simplify lets think of a neural network for making a decision whether to hire an employee or not
* Lets consider three weights
* experince = 0.6
* skills = 0.3
* communication = 0.1
* Activation function: hire a canditate if the score is > 5
* Example 1: 5 * 0.6 + 4 * 0.3 + 3 * 0.1 = 3 + 1.2 + 0.3 = 4.5
* Manager has a gut feeling that this candidate is good so he adds a bias of `1` => 4.5 + 1 = 5.5
* formula w1x1+ w2x2 + ...+wnxn + b* While training a neural network it will predict wrong values and then the cost (difference between predicted and actual value) then back propagation happens which adjusts weights and biases
* For Basics of NLP refer the [notes uploaded](https://directai.blog/2025/02/12/gen-ai-notes-12-02-2025/)
* Three terms to research
* stop words: and or is…
* Stemming: Removes suffixes
* studies => studi
* running => run
* lemmatization: goes to dictionary equivalent
* studies => study
* tokens: word is broken down into multiple tokens
