Deperrois, Nicolas Rouben Pascal (2023). Learning to Dream, Dreaming to Learn. (Thesis). Universität Bern, Bern
|
Text
23deperrois_nrp.pdf - Thesis Available under License Creative Commons: Attribution (CC-BY 4.0). Download (6MB) | Preview |
Abstract
The importance of sleep for healthy brain function is widely acknowledged. However, it remains mysterious how the sleeping brain, disconnected from the outside world and plunged into the fantastic experiences of dreams, is actively learning. A main feature of dreams is the generation of new realistic sensory experiences in absence of external input, from the combination of diverse memory elements. How do cortical networks host the generation of these sensory experiences during sleep? What function could these generated experiences serve? In this thesis, we attempt to answer these questions using an original, computational approach inspired by modern artificial intelligence. In light of existing cognitive theories and experimental data, we suggest that cortical networks implement a generative model of the sensorium that is systematically optimized during wakefulness and sleep states. By performing network simulations on datasets of natural images, our results not only propose potential mechanisms for dream generation during sleep states, but suggest that dreaming is an essential feature for learning semantic representations throughout mammalian development.
| Item Type: | Thesis |
|---|---|
| Dissertation Type: | Single |
| Date of Defense: | 24 January 2023 |
| Subjects: | 100 Philosophy > 150 Psychology 600 Technology > 610 Medicine & health |
| Institute / Center: | 04 Faculty of Medicine > Pre-clinic Human Medicine > Institute of Physiology |
| Depositing User: | Sarah Stalder |
| Date Deposited: | 02 May 2023 08:37 |
| Last Modified: | 24 Jan 2024 23:25 |
| URI: | https://boristheses.unibe.ch/id/eprint/4258 |
Actions (login required)
 |
View Item |