Tweeted By @hardmaru

Deep ConvNets trained to classify images of objects while also trained to have “representations that resemble those observed in neural recordings from monkey visual cortex” have better generalization abilities and are also more robust to label corruption.https://t.co/GoxnaD3uos https://t.co/53qJcVHlhY pic.twitter.com/FWumP1hPYD

— hardmaru (@hardmaru) August 11, 2020