3 ms·
> Not to be negative about the research, but I’m pretty sure I could have surmised the pigs’ emotional state in many of these situations (castration?!) without
by root_axis 5y ago
> Not to be negative about the research, but I’m pretty sure I could have surmised the pigs’ emotional state in many of these situations (castration?!) without the benefit of a translator!
That's the point. The obvious emotional states are necessary in order to correctly label the data for training. The interesting part is to then use the trained model to identify emotional states in situations where it isn't so obvious.
- kjkjadksj 5y agoIf you train on obvious cases its not going to be a guarantee that your model is robust on less obvious cases.
- pishpash 5y agoGarbage in garbage out. How do we know the emotional states are what they are? Better to use as label something objectively measurable but costly to observe like some internal brain or chemical state of the pigs.
- sacrosancty 5y agoHow about just named after the situations they were observed in? Then when you give it a new pig sound, the model can classify it and we can decide how to respond to "slaughterhouse=0.8, castration=0.2" or whatever. I'm especially suspicious of fighting being negative. Humans can get pleasure from fighting. Why not pigs?