I just took my prescription pills. One of these pills is a soft-gel that I found to work the best when I chew it before swallowing (probly works better in the upper digestive system than lower). Maybe someday, we'll have nutrition pills that require chewing like real food. And we'll just chew on pills for sustenance.
Anyways, I wanted to post something I was discussing with a really good friend last night. The models and algorithms that we were interested in require analysis and optimization in these properties:
- Latency and Throughput (Needs to respond fast enough, and supply solutions fast continuously)
- Relevance. (This is often ignored because most of the time designers of algorithms will not deviate with from the problem)
- Performance. (Actual performance in the sense of accuracy, coverage, generalizability, stability and robustness, and min-max properties, etc.)
- Cost: how much is it goina cost to apply an algorithm satisfying the above three types of specifications. This should also include implementation cost. Something like TCO, possibly minus the O at some point. :(
- Long term performance: What it will do in the long term. This includes every thing else that is more global than what is required to satisfy the above four and contain things like environmental issues, game theoretic problems, and the likes.
This may seem completely out of context to everybody, but that one friend who's email system suddenly all barfed, here they are. the properties that I want to keep track of.
goina drink more now. wanted to get that down on electrons before I pass out.
No comments:
Post a Comment