# Artificial neural networks: The idea

Source page: https://www.materialinsights.org/science/artificial_neurons/sci-idea/

#### In plain words

The idea explained without jargon

A nerve cell adds up the signals reaching it from other cells and fires if the total is big enough. An artificial neuron does the same sum: it multiplies each input by a number called a weight, adds the results and outputs 1 if the total passes a threshold, otherwise 0. McCulloch and Pitts showed that networks of such all-or-none units can carry out logical operations ^[1]. In Rosenblatt's perceptron the weights are not set by hand: the connections change with experience, and reinforcement can help or hinder that change ^[3]. One layer can only split inputs with a straight line, so it cannot solve XOR (one or the other, but not both); adding a hidden layer removes that limit ^[5].

as of 2026-11-12

#### Key equation

The equation that sums it up, and what each symbol means

**Equation**
: y = 1 if w1 x1 + w2 x2 +... + wn xn >= theta, otherwise y = 0

**Symbols**
: {"y":"the neuron's output, 1 for firing and 0 for silent","x1... xn":"the inputs, from other neurons or from the data","w1... wn":"the weights: how strongly each input counts, positive or negative","theta":"the threshold the weighted total must reach"}

**Note**
: This is the all-or-none threshold unit of McCulloch and Pitts ^[1]. Modern networks replace the sharp threshold with a smooth curve so that training by gradient methods is possible. Learning means changing the weights, not the rule.

as of 2026-11-12

#### An analogy

A picture from everyday life that makes the idea easier to hold

A neuron is like a member of a panel who votes only if the weighted opinions of the people advising it add up to enough. Learning is deciding, after each outcome, how much to trust each adviser. The picture stops being true because real neurons fire in timed pulses and come in many kinds, and because a single panel member cannot solve every problem: tasks like XOR need a second row of members in between.

as of 2026-11-12

#### Sources

1. [A logical calculus of the ideas immanent in nervous activity](https://link.springer.com/doi/10.1007/BF02478259) The Bulletin of Mathematical Biophysics (W. S. McCulloch and W. Pitts), Springer, 1943, vol. 5, December 1943
2. [The Perceptron: A Probabilistic Model for Information Storage and Organization in the Brain](https://homepages.math.uic.edu/~lreyzin/papers/rosenblatt58.pdf) Psychological Review (F. Rosenblatt), 1958
3. [Cornell Virtual Workshop > Scientific Machine Learning (SciML) > Multi-Layer Perceptron > The Need for Depth - The XOR Problem: A Historical Turning Point](https://cvw.cac.cornell.edu/SciML/mlp/xor-problem) Cornell University Center for Advanced Computing

## Other sections

- [Summary](https://www.materialinsights.org/science/artificial_neurons/index.md)
- [Discovery](https://www.materialinsights.org/science/artificial_neurons/sci-discovery/index.md)
- [Technologies](https://www.materialinsights.org/science/artificial_neurons/sci-technologies/index.md)
