What Is a Neural Network?
Short Answer
A neural network is a machine learning model made of layers of simple connected units called neurons. Each connection has a weight, a number the network adjusts during training until its outputs match the examples. Stacked in many layers, neural networks are the basis of deep learning.
- Taught in lesson 73
Why It Matters
Neural networks are the building block of deep learning and of every large language model. Knowing roughly how they learn helps you understand training, fine-tuning, GPUs and why models need so much data, even if you never build one from scratch.
How It Works
Each neuron multiplies its inputs by weights, adds them up and passes the result through an activation function. Training uses backpropagation and gradient descent: the network measures its error, works out how each weight contributed, and nudges every weight to reduce the error.
The idea was loosely inspired by the brain, but modern neural networks are mathematics, not biology. Large language models are neural networks with billions of weights, also called parameters.
The idea was loosely inspired by neurons in the brain, but modern neural networks are mathematics, not biology. Each unit computes a weighted sum of its inputs and applies a simple function. What makes networks powerful is scale and training: millions or billions of weights, adjusted step by step to reduce the error on many examples. Different arrangements of units suit different data, such as convolutional networks for images and transformers for language.
Neural Network in Practice
Everyday Example
A network that reads handwritten digits takes the pixels of an image as input and outputs which digit, 0 to 9, it most likely shows.
Industry Example: Pharma
Neural networks read microscope images of cells to measure how a compound changes them, a step in screening thousands of candidate drugs.
Where You Meet It
- Any explanation of deep learning or language models.
- PyTorch and TensorFlow code.
- Discussions of model size, measured in parameters (weights).
- Diagrams of layers and arrows in AI courses and papers.
Key Ideas Around It
- Weight
- A number on a connection, adjusted during training.
- Activation function
- A simple function that lets the network learn non-linear patterns.
- Backpropagation
- How the network works out which weights to change.
- Parameter
- Another word for a learned weight; model size is counted in parameters.
A Common Mix-Up
A neural network does not think like a brain. The name is a historical metaphor; the network is a large mathematical function tuned from data.
Check YourselfWhat changes inside a neural network when it learns?
Its weights: training adjusts the numbers on its connections so its outputs get closer to the right answers.
Learn It on ai2hired
S7 · #73A Neural Network From Scratch
Coming in Season 7, Deep Learning and Transformers. Every ai2hired lesson is free on YouTube.
Questions People Ask
What is a parameter in a neural network?
A parameter is one of the numbers the network learns during training, mostly the weights of its connections. Model size is often given as a parameter count.
Is a neural network like a human brain?
Only loosely. It borrows the idea of many simple connected units, but it learns and works very differently from biological neurons.
Related Terms
Learn AI From Zero to Hired
Free lessons on YouTube, each one built from what real AI job ads ask for.