The example begins with picking a particular boolean function of, say, 5-10 variables. The goal is to find a circuit that computes this function. We need an evaluation method, stopping criteria, a model space of circuits (schema and dimensions).
One interesting thing about this example is that, by collecting data on why a particular model fails, we might be able to use that info to more efficiently explore the search space (loss function result being a vector, not just a scalar, in a sense)
The example begins with picking a particular boolean function of, say, 5-10 variables. The goal is to find a circuit that computes this function. We need an evaluation method, stopping criteria, a model space of circuits (schema and dimensions).
One interesting thing about this example is that, by collecting data on why a particular model fails, we might be able to use that info to more efficiently explore the search space (loss function result being a vector, not just a scalar, in a sense)