An interesting theory on the meaning of dreams.

We know broadly what happens in the brain when we dream. We are just not sure why. Brain activity during dreams are not dissimilar to those when awake. But there are additionally biochemical processes that inhibit associated behaviour. This is so that dreaming people generally do not wander around acting out their dreams. Hoel argues against the popular notion that dreams occur either to consolidate or to replay memories. He suggests instead that they prevent overfitting. This is a problem in deep learning whereby the neural network becomes overly fine-tuned to the training data. The brain’s neuronal networks have some similarities to artificial neural networks. So, it is possible that the brain also overfits its experience in forming learning and memory. The answer to overfitting in deep learning is to introduce stochasticity and noise to the network. Perhaps dreams, in all their weirdness, perform the same task for brains.

Link to article:

You may also like to browse other neuroscience articles: