Vincent brought up a subtle but important point that for semi-supervised learning to be practically useful, it has to work on both low data regime and high data regime. And this is what we begin to see in these two papers.
Prikaži ovu nit
Nice blog post titled "The Quiet Semi-Supervised Revolution" by Vincent Vanhoucke. It discusses two related works by the Google Brain team: Unsupervised Data Augmentation and MixMatch. https://towardsdatascience.com/the-quiet-semi-supervised-revolution-edec1e9ad8c …pic.twitter.com/bbDxaF6vep
-
-
-
Links to the mentioned papers. MixMatch: https://arxiv.org/abs/1905.02249 Unsupervised Data Augmentation: https://arxiv.org/abs/1904.12848
Prikaži ovu nit -
Key idea in these two papers is to ensure prediction(x) = prediction(x + noise) , where x is an unlabeled example. People have tried all kind of noise, e.g., Gaussian noise, adversarial noise etc. But it looks like data augmentation noise is the real winner.pic.twitter.com/ClUOqbvEsj
Prikaži ovu nit -
To add to Vincent's point above, new findings also include: 1. The method is general (works well for images & texts). 2. The method works well on top of transfer learning (e.g., BERT). You can find these results in Unsupervised Data Augmentation paper: https://arxiv.org/abs/1904.12848
Prikaži ovu nit
Kraj razgovora
Novi razgovor -
-
-
Your paper "Semi-Supervised Sequence Modeling with Cross-View Training" is also on the same topic with a focus on sequence labeling tasks.
-
Yes, you're correct. Thank you for bringing this up.
Kraj razgovora
Novi razgovor -
-
-
Semi-parametric methods like Cox regression are useful because they relax assumptions and trade reliability for accuracy. This is the first time I've seen it claimed a more specific model with more assumptions should be expected to be less accurate.
-
I'll have to see if there's a toy example. Shouldn't be too difficult to whip up a small simulation study to check.
- Još 2 druga odgovora
Novi razgovor -
-
-
Thank you.
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
Very interested to discuss data augmentation for NER...if you pass by TC3 in Sunnyvale, let me know for a short discussion. Coffee on me :-)
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
Čini se da učitavanje traje već neko vrijeme.
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.