Lugh@futurology.todayM to Futurology@futurology.todayEnglish · 4 个月前Even with poorly labeled training data, new research shows AI outperforming Radiologists in diagnosing rarely encountered conditions.www.openicpsr.orgexternal-linkmessage-square7fedilinkarrow-up121arrow-down10
arrow-up121arrow-down1external-linkEven with poorly labeled training data, new research shows AI outperforming Radiologists in diagnosing rarely encountered conditions.www.openicpsr.orgLugh@futurology.todayM to Futurology@futurology.todayEnglish · 4 个月前message-square7fedilink
minus-squarethreelonmusketeers@sh.itjust.workslinkfedilinkEnglisharrow-up3·edit-24 个月前If the conditions are rare, and training date is poorly labeled, isn’t there a danger that these models could be overfitting?
If the conditions are rare, and training date is poorly labeled, isn’t there a danger that these models could be overfitting?