Profile cover photo
Profile photo
Ran Manor
Piled higher and deeper
Piled higher and deeper
About
Ran's posts

Post has attachment
אני מתחילון, מה הסטטוס שלך? הנה הקוד שלי GTMVCEJ, שיעניק לך ₪15 לנסיעתך הראשונה. http://invite-il.gett.com/code/GTMVCEJ
Photo

tl;dr - noisy labels in neural networks

I have some data I'm classifying using a neural network. I've noticed that the training labels don't always match the training samples (although most of the time they do match).
Is there anyway to incorporate this in the neural network?
Thanks.

tl;dr - how do I check which features the strongest?

I built a neural network that has two inputs, each of them goes through several layers, and after that the final features are concatenated to one feature vector which is classified in the output layer.

I want to see, for a single sample, which input contributed most to the classification outcome. I thought about looking at the norms of the final features of each network, but the problem is that one input produces 4096 features and the other one produces 40 features, so they are not really comparable, as the norm of the largest vector is always larger.

Anyone has an idea about this?
Thanks!

Post has attachment

Post has attachment

Post has attachment
Google releases its latest machine learning system.

Post has attachment

Post has attachment
Use my Uber promo code, erpac, and get ₪30 off your first Uber ride. Redeem it at https://www.uber.com/invite/erpac

Post has attachment

Post has attachment
Wait while more posts are being loaded