Abstract: An online system receives contextual information for a goal-oriented environment at a current time and generates Q-value predictions that indicate likelihoods that one or more participants will reach the desired goal. The Q-value for a current time may also be interpreted as the value of the actions taken at the current time with respect to the desired goal. The online system generates Q-value predictions for a current time by applying an approximator network to the contextual information for the current time. In one instance, the approximator network is a machine learning model neural network model trained by a reinforcement learning process. The reinforcement process allows the approximator network to incrementally update the Q-value predictions given new information throughout time, and results in a more computationally efficient training process compared to other types of supervised or unsupervised machine learning model processes.

Abstract: A plurality of modular tiles can be connected to form a writing surface. The modular tiles are shaped to enable multiple tiles to be connected to form a continuous surface (e.g., hexagonal, rectangular, etc.). Each modular tile has a writing element and a friction element opposite the writing surface to prevent the modular tile form slipping when placed on a horizontal surface. The modular tiles may also have a coupling mechanism, such as magnets with appropriate polarization, allowing the modular tiles to be coupled to one another to form a large surface and then be detached for simpler storage of the modular tiles. The modular tiles may further include lights, displays, or additional electronics to allow a connected computing device to display on the writing surface.