Re:More Data Is Not Better and Machine Learning Is a Grind
by gnulinux on Wednesday January 16, 2019 @12:15AM
(#18918415)
This is not surprising because if image A is an apple, its mirror image A' is also an apple. Adding A' to your dataset is just plain ol' regularization. It is meant to lower bias, to prevent overfitting.
[ Parent
| on HN ]
Re:More Data Is Not Better and Machine Learning Is a Grind
by jzwinck on Wednesday January 16, 2019 @05:45AM
(#18919742)
It will also make the flag of Côte d’Ivoire be perceived as that of Ireland.
Re:More Data Is Not Better and Machine Learning Is a Grind
by Topolomancer on Wednesday January 16, 2019 @02:01AM
(#18918824)
It's not surprising, but highly necessary because you might otherwise learn too little about structural features, and too much about unnecessary things such as the background colour of some objects.