Abstract: A first set of features associated with a neural network are parameterized. A decision tree is generated from the first set of features. One or more adjustments for the neural network are received at the decision tree. A second set of features associated with the adjustments at the decision tree are parameterized. The parameterized first and second set of features are combined into a plurality of parameters. From the plurality, an adjusted neural network is generated.

Abstract: A first set of features associated with a neural network are parameterized. A decision tree is generated from the first set of features. One or more adjustments for the neural network are received at the decision tree. A second set of features associated with the adjustments at the decision tree are parameterized. The parameterized first and second set of features are combined into a plurality of parameters. From the plurality, an adjusted neural network is generated.

Abstract: A building-information management (BIM) system receives a set of wind vectors that identify wind speeds and directions at the surfaces of physical structures within a volume of air. The system stratifies the volume into vertically stacked layers that each contains similar wind vectors. Each layer is then divided horizontally into rectangular cells that each at least partially encloses a cluster of the structures. The system infers cell-specific relationships that relate physical attributes of the enclosed structures with the cell's incoming and outgoing airflows, from which the system builds a wind-propagation model for each cell. These cell-specific models are combined into a regional wind-propagation model that predicts how winds entering the volume will be propagated as outgoing wind vectors. The regional model is used by downstream modules of the BIM system to account for airflow effects when designing a planned building.

Abstract: A building-information management (BIM) system receives a set of wind vectors that identify wind speeds and directions at the surfaces of physical structures within a volume of air. The system stratifies the volume into vertically stacked layers that each contains similar wind vectors. Each layer is then divided horizontally into rectangular cells that each at least partially encloses a cluster of the structures. The system infers cell-specific relationships that relate physical attributes of the enclosed structures with the cell's incoming and outgoing airflows, from which the system builds a wind-propagation model for each cell. These cell-specific models are combined into a regional wind-propagation model that predicts how winds entering the volume will be propagated as outgoing wind vectors. The regional model is used by downstream modules of the BIM system to account for airflow effects when designing a planned building.