Pattern recognition is a poor fit for fat long tails of exceptions. It's not just the data, it's the model category. You could say that ML models are stereotyping machines.
Show this thread
A common, important way in which many real-world ML models are biased (beyond data bias) is that they privilege "canonical" samples and incorrectly process edge cases. This is the main failure mode of recommender systems, for instance. Not a matter of data representativeness.https://twitter.com/yoavgo/status/1274826491436687360 …
-
-
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Almost any ML method could be made statistically robust, ie resistant to outliers.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Is this the same thing as high "bias" in a "bias vs variance" sense, or are you referring to something different? It seems like what you are describing is a system too "simple" (in some way) to capture all the underlying dynamics of the thing it's trying to model.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Is that not an issue at some point to treat scale effect? Should you not renormalize using ref value? A little bit like in physics / engineering when you non dimensionalise.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
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.