4 ms·
If you're interested in doing more ML work, here's an interesting thought experiment to guide your research towards an idea: ML is currently widely used for thi
by structural 9y ago
If you're interested in doing more ML work, here's an interesting thought experiment to guide your research towards an idea: ML is currently widely used for things like speech processing and synthesis, music, as well as vision. All of this is sensing and reacting to stimuli at various frequencies. But there is much more... So from the lowest to highest practical frequencies, what do we use each band for, and what would it mean to see in it?
Some examples:
- Single digit Hertz: seismic observations. Could you build a better earthquake predictor? How about an active control system for shifting building loads during an earthquake?
- kilohertz: Lightning detection systems.
- kilohertz to low megahertz: traditional FM/AM radio. Is there a better way?
- gigahertz: think your good old-fashioned microwave oven. What if it could automatically learn what it's cooking and how to best heat it evenly?
- gigahertz: could you build a smarter Wi-Fi network (or cell network) if devices reported their location when they didn't have coverage and the network could automatically adjust?
- tens of gigahertz: we now have object sensing systems in cars. what other parts of life might that kind of technology be useful for?
- infrared: what happens when a ML-based vision system wants to see in the dark?
- above the visual spectrum: ML + medical imaging (x-rays, other scans? what could be learned?)
There are lots of fun research questions to be answered and new technology yet developed. The wonderful part about doing some really structured research along these lines is that you might have to ask about "who knows about this" and run into the PhD student who's working in this area -today- but isn't yet developing a product. There is a ton of university research that could lead you towards an interesting new product: the idea doesn't have to be your own, or even come from the typical "ideas" person.