If you have a signal where a certain portion of points are noise, and a certain portion of points are "interesting", what's the "best" way to determine which points are "interesting" as they come in over time?
-
-
-
Iseems it could be treated as a time series classification problem (https://arxiv.org/pdf/1910.13051.pdf …) , also it would depend on how you frame the model, arch with attention are good at focusing and handle signal data, you would just need to monitor de attention layer
1 reply 0 retweets 1 like -
Nice paper and example of just how complex this can get if accuracy is critical verses say real-time trading where speed/cost is a factor and bets can be corrected/adjusted ( PID loop ) - good enough vs perfect.
1 reply 0 retweets 2 likes -
yup and also going to your previous point how much "we" gloat on using raw data (which increases complexity) and if we would complement with "expert" data, things would work faster/better
1 reply 0 retweets 1 like
Yep. Dimensionality reduction - if you need to be in production tomorrow verses you have an entire PhD's worth of bandwidth then mileage may vary 
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.