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 months

Recordings per month over the last 12 months — 1 in all, peaking in Jul 2026 with 1.

Appearances

newest first · ▶ plays the moment

Talking Techniques · AI & Antibodies miniseries | Nanobody thermostability prediction and the great data challenge · 6 Jul 2026 podcast

Still enough to to learn something out of it.

And how we did that is that we made sure that we gonna leverage other tools.

to improve uh training.

So we use pre-trained large language model on proteins that allow to um through this transfer learning allow to um increase the accuracy of the prediction and we also develop a robust pipeline to make sure that because it's a lack of um standardized data, you have a lot of noise.

And then based on how you're gonna split the training data set, you might um

poorly select the hyperparameters during your search or you you might have a biased performance assessment.

So we did like this repeated cross-validation.

We made sure the the training pipeline was robust to the noise of the data to have a more representative and true

Assessment.

And this is you need to be careful when you have a small data set.

Yeah, so accuracy.

So you need to define first accuracy.

What's accuracy and on on which kind of uh

test set you're gonna look at.

So we uh kept uh held that test set for the training.

I mean during my the development of the of the of nanomelt, we measured more and more uh

permostability in the lab and more people publish more.

And so at the end of the project, we had 80 more data points Unique that the model has never seen.

Um and we make sure that uh this data set is uh

diverse and is far away from the training set.

Showing 41–60 of 346 · page 3 of 18