What Is Hebbian Theory?

Most people think that reading and writing are the same things. They're not. Reading transfers information from one place to another, while writing creates something new. That's why when you read something, you understand it, but when you write something, you must make it up yourself. That's also why we need Hebbian Theory—to help us understand how synapses change over time and what factors contribute to their plasticity (or lack thereof). This theory was first proposed by the Canadian psychologist and neuroscientist Donald O. Hebb in 1949. It remains a widely used model for understanding how we acquire new knowledge and form new memories. Hebbian theory models many aspects of artificial neural networks, such as recognition, pattern detection, and other functions dependent on dynamic connection strengths between neurons. In modern artificial neural networks, algorithms can rework the weights of neural connections. Professionals sometimes talk about "Hebb's rule," which describes how these connections work and change. The element of the appeal of Hebbian theory is that by changing neural weights and associations, engineers can get different results from sophisticated artificial neural networks. The basic premise behind Hebbian learning is simple: if two neurons are connected (in this case, by an axon), then whenever one neuron fires, the other neuron will have a higher probability of firing. This means that if you have an input neuron that fires when a person sees something blue and an output neuron that fires when people see something red, then whenever someone sees something blue, their brain will activate both neurons at once--and vice versa for red objects! In this way, artificial neural networks can be trained to recognize patterns in data--they can learn what signals mean based on patterns in their inputs and outputs over time.