Larry DialSETS P6: Power Mean Aggregations. More Hyperparameters!Tree search algorithms may backpropagate information from downstream nodes. To keep memory sizes reasonable, this backpropagated…6d ago6d ago
Larry DialSETS P5: Breaking the Curse of Dimensionality with Correlated Uncertainty in Value EstimatesThe search space for sequential decision-making problems grows exponentially with the number of decisions. This means that solving two…6d ago6d ago
Larry DialSETS P4: Rethinking Scalar Optimization Objectives2 Player Zero Sum games tend to have a binary Win/Loss objective, and so it is natural to work with values in the range of 0 to 1…Sep 21Sep 21
Larry DialSETS P3: Deep Rollout Tree SearchIn the same way that MCTS mirrors how humans might search over options in a chess game, the proposed algorithm Deep Rollout Tree Search…Sep 21Sep 21
Larry DialSETS P2: AlphaZero as a starting pointThis project originally started as the idea that I would directly implement Neural MCTS, as done in AlphaZero, to the problem of training a…Sep 20Sep 20
Larry DialSearching for Efficient Tree Search (SETS): P1 MotivationLately I have been exploring new approaches to tree search for sequential decision making in large deterministic environments with many…Sep 18Sep 18
Larry DialApproaching RTS Game Balance- thoughts from a random guyMuch is said on the topic of RTS balance. Few travelers who enter the land of cheese, whine, and salt ever are found again- at least in any…Jan 26Jan 26
Larry DialBell’s Theorem Analogy: Bob at the CasinoBob approaches a new game at the casino. The dealer has arranged playing cards facedown along the perimeter of a circular table. The dealer…Nov 1, 2022Nov 1, 2022
