Wednesday, December 18, 2013

preliminary ideas for a competition of low power

I think the competition should be at the system level, something most people can understand. This can create more excitement and wider participation, like what Deep Blue and Watson did. If it is low-level, such as "being able to use only 1e-20J doing double-precision floating-point division", most people will not understand.

These are some preliminary ideas: 

We can create some types of "recognition-based benchmarks". For example, give 1,000 + 1 images. Among the 1,000 images, one of them is the closest to that particular image. The "closest" will be defined by a committee but should be commonly agreed by most people. If 1,000 are too few, we can easily increase that to much more. ILSVRC ( has a lot of images.

The participants are divided into two groups: 

  1. customer hardware
  2. commodity hardware. 
We need to define the boundary of the two groups. 

We may further divide the groups to 
  1. with network connections (can offload computation).
  2. without network connections (cannot offload computation).
We need to specify well-defined interfaces for measuring power and energy. We may also specify the minimum amount of ambient light so that participants can use solar power, if they choose so.

No comments:

Post a Comment