Click "Evaluate" below to run the computation. This SAGE cell adapts utilizes the backpropagation algorithm from https://mathrepo.mis.mpg.de/PolynomialNeuralNetworks/ to compute the dimension of neurovariety for a given architecture (d0,d1,...,dL) and monomial activation function. For a database containing computational results, please refer to https://github.com/daokevin06/Expressivity_Thresholds/tree/main/data/processed The database was built using a CUDA sped-up version of the following algorithm.