What is your Scientific motivation?
What are some of the limitations you face in this problem? For example, is it too slow right now? Is it very tedious to iterate through different configuration?
Is there some criterion a solution has to fulfill? For example, it has to be able to process n examples (throughput), or it has to be accurate down to some precision limit.
Is uncertainty important?
Is there some smoothness condition you think would matter?
What's the dimension of your input and output? Is it constant or variable?
How much data do you have for training? Hundreds of GB? MB? or no data at all?
Do you expect the underlying process to be stochastic?
How many data channels you have?
Provide a short example to access and construct the test problem.
Current method benchmark.
Suggest methods and why they might help.
Brief list of tasks for building the network for beating the benchmark/production
failure mode