Bizarre dreams, according to specialists at the Human Brain Project, may help your brain learn more effectively.
The National Sleep Foundation estimates that we dream four to six times every night on average. However, because we forget over 95% of our dreams, you will only recall a couple per month.
According to a new study from the University of Bern in Switzerland, dreams, particularly those that appear authentic but are later discovered to be aberrant, assist our brain in learning and extracting general ideas from previous experiences. Using machine learning-inspired methodologies and brain modeling, the study, which was done as part of the Human Brain Project and published in eLife, proposes a novel hypothesis on the significance of dreams.
Sleep and dreams have long been recognized as important in learning and memory, and the impact that a single restless night may have on our cognition has been extensively documented. However, there isn’t a theory that connects experience accumulation, concept generalization, and creativity.
Non-REM sleep, in which the brain “replays” the sensory stimulation encountered while awake, and REM sleep, in which spontaneous bursts of strong brain activity form vivid dreams, are the two types of sleep phases that we usually experience during sleep, alternating one after the other. The researchers used brain cortex simulations to simulate how different sleep phases affect learning. They were inspired by a machine learning approach known as Generative Adversarial Networks (GANs) to add an element of unusualness to manufactured dreams.
Two neural networks compete to produce new data from the same dataset, in this case a collection of simple photos of objects and animals, in GANs. This technique creates fresh artificial visuals that appear to a human viewer to be realistic on the surface. After that, the researchers simulated the brain in three different states: awake, non-REM sleep, and REM sleep. The model is shown photographs of boats, vehicles, dogs, and other objects while awake. The model replicates the sensory inputs with various occlusions in non-REM sleep.
Through the GANs, REM sleep generates new sensory inputs, resulting in distorted but lifelike copies and mixtures of boats, vehicles, pets, and other objects. A simple classifier analyzes how quickly the identification of the item can be read from the cortical representations to measure the model’s performance.
“Non-REM and REM dreams become more realistic as our model learns,” explains Jakob Jordan, senior author, and leader of the research team. “While non-REM dreams resemble waking experiences quite closely, REM dreams tend to creatively combine these experiences.”
The accuracy of the classifier reduced when the REM sleep period was inhibited in the model, or when these dreams were rendered less imaginative. These representations tended to be more susceptible to sensory perturbations when the NREM sleep period was abolished (here, occlusions). Wakefulness, non-REM sleep, and REM sleep all appear to have complimentary functions in learning: experiencing the stimuli, cementing that experience, and discovering semantic concepts, according to this research.
“We think these findings suggest a simple evolutionary role for dreams, without interpreting their exact meaning,” says Deperrois. “It shouldn’t be surprising that dreams are bizarre: this bizarreness serves a purpose. The next time you’re having crazy dreams, maybe don’t try to find a deeper meaning – your brain may be simply organizing your experiences.”