Aubin Ramon
speaker
346 appearances
1 recordings
1 series
first heard Jul 2026
last heard 6 Jul
Aubin Ramon’s voice in public audio — every appearance, attributed to the second.
Trend
recordings per month · last 12 monthsRecordings per month over the last 12 months — 1 in all, peaking in Jul 2026 with 1.
Appearances
I mean it's about folding.
Um and we try to incorporate some structure awareness into a model.
Um
But because the quality of the structure we had and I was that wasn't working.
So I tried to have this big graph neural network.
um based on the structure.
Um and then the model became very, very big and that was just overfitting into uh my data set, my training data set.
So my advice, my piece of advice here is um sometime go for the simplest model.
Don't try to go for something big because deep learning works for deep data, so for a lot of data.
So if you have a limited data set, go for um
um a more simple model and maybe you can start by training your model on this dataset on a simple baseline.
So then you know that what the amount of overfitting you can get with just a simple model.
And then from that you can start to implement
um incrementally some new features, some more elegant, um loss functions, some things on the top, um, maybe an extra layer that do gonna do something, and keep comparing with the the um overfitting baseline, trying to make sure that you're not gonna fit overfit more.
Uh
This is what I I will do, yeah.
Fantastic.
Yeah, so I mean one year is not that much for for for research.
Um
Yes, so I mean the first
Showing 261–280 of 346 · page 14 of 18
← Previous
Next →