Previous | Next --- Slide 35 of 41
Back to Lecture Thumbnails
kipper

There's a lot of interest in trying to make neural networks more efficient from the energy perspective, given that these networks tend to be really large. As the slide shows, just storing and reading so many parameters can be a really energy sink for the applications. Some of the ways people are investigating to do so are discussed on the following slides, including network compression and running on GPUs.