Neural Network

What a Neural Network Is

A neural network is a type of machine learning model made up of layers of simple mathematical units called nodes (or neurons). These nodes pass values to each other, with each connection having a weight that determines how much influence one node has on the next.

The name comes from a loose analogy to biological brains, where neurons fire signals to each other. The mathematical reality is simpler — it's matrix multiplication all the way down — but the analogy helps with intuition.

How the Layers Work

A typical neural network has three kinds of layers:

• Input layer — receives the raw data (pixels in an image, tokens in text)
• Hidden layers — one or more intermediate layers that learn increasingly abstract representations of the data
• Output layer — produces the final result (a label, a probability, the next token)

"Deep learning" simply means neural networks with many hidden layers — "deep" refers to the depth of the network. More layers allow the network to learn more complex patterns.

How Training Works

During training, the network makes predictions and compares them to correct answers. When it's wrong, an algorithm called backpropagation adjusts the weights slightly to reduce the error. After millions of examples and billions of these tiny adjustments, the weights settle into values that produce accurate predictions.

This is what "a trained model" means — a network whose weights have been optimized through training.

Connection to LLMs

Large Language Models are neural networks — specifically, they use the transformer architecture, which is a particular design of neural network optimized for processing sequences of text. The "large" in LLM refers to having hundreds of billions of parameters (weights).