I've added the Inception v3 model to our model collection:https://github.com/fchollet/deep-learning-models …
-
-
Replying to @fchollet
minor point: if someone doesn't want to use pretrained models, the default is no weight reg, which is suboptimal
1 reply 0 retweets 0 likes -
for example, the keras resnet has no weight reg, when the FB team trained it with 0.0001 l2 reg on bias + weights.
1 reply 0 retweets 0 likes -
Replying to @F_Vaggi
L2/L1 regularization is basically always harmful on this type of model. Regularization should be done via dropout/noise/etc.
1 reply 0 retweets 2 likes -
Replying to @fchollet
The Facebook AI team trained resnets with l2 reg, and so did the original MS research paper:https://github.com/facebook/fb.resnet.torch …
2 replies 0 retweets 0 likes
yes, I know. You can verify by yourself that it was a mistake (though not as bad a mistake as using Torch instead of Caffe/TF ;-)).
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.