A research workstation is training three deep learning models simultaneously: a natural language transformer, a convolutional neural network for image processing, and a reinforcement learning simulation.
The combined RAM requirement for these three concurrent training processes is 48 GB, but the workstation is only equipped with 24 GB of physical RAM.
Despite this physical limitation, all three models continue to train successfully without throwing "Out of Memory" system errors.
Explain how the operating system utilizes virtual memory to manage this memory deficit and keep all three processes running.