Skip to content

Latest commit

 

History

History
36 lines (18 loc) · 1000 Bytes

Template.md

File metadata and controls

36 lines (18 loc) · 1000 Bytes

Question

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?

Dataset

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 status

Current method benchmark.

Potential method

Suggest methods and why they might help.

Brief list of tasks for building the network for beating the benchmark/production

failure mode