Negative Data Augmentation

Показать описание

This video explains Negative Data Augmentation, a strategy for using label-corrupting, rather than label-preserving transformations in Deep Learning. The authors test this framework for training GANs and for Contrastive Learning such as CPC and MoCo. I think this is a really exciting direction for Data Augmentation and overcoming the challenge of learning from limited labeled data, I hope you find this video useful!

Content Links:

Chapters
0:00 Beginning
0:55 Semantically-Preserving Transformations
1:44 OOD Augmentations
3:25 NDA Strategy
6:00 Over-Generalization
7:18 Integration in GANs
8:00 Integration in Contrastive Learning
8:40 GAN Results
11:00 Contrastive Learning Results
12:18 Dark Matter - Energy-Based Learning
13:30 The Diff that makes a Diff

Connor Shorten

Рекомендации по теме

Комментарии

I can see this having a benefit in music generation - shuffling parts of the song around or splicing to other songs...

terryr

Good to hear u back on regular uploads :)

dawwdd

great idea... it resembles the gradient reversal idea of DANN.

nikre

Interesting. But I'm not sure it's what we need. I mean, is it wrong to classify an image as "dog" when there's only half his face on the image? What I see in the dog example is 4 sub-images, and each *should* be labelled "dog" if tested independently.
My point is actually related to the concept of equivariance (as opposed to invariance).

Maybe the jigsaw transformation should have more than 4 parts to really destroy the high-level structure?

Ceelvain

I m interested if someone know where i could find some code for those augmentations :)

yoannfleytoux

I am curious if anyone has any ideas how to add these negative examples to a VAE (something like Jukebox)

terryr

Negative Data Augmentation

Negative Data Augmentation

negative data augmentation

Rethinking Rotation in Self-Supervised Contrastive Learning: Adaptive Positive or Negative Data Aug

Domain Generalization for Face Anti Spoofing via Negative Data Augmentation

Data Augmentation for Machine Learning, Best Practices Explained with Examples!

New Data Augmentation Technique - Dealing with Imbalanced datasets

Machine Learning for COVID-19: Solving Data Deficiency via Data Augmentation

GANs for Data Augmentation

Text Augmentation To Boost NLP Model Performance (Part 1/3)

Data Augmentation for Graph Neural Networks

Data Augmentation using Pre-trained Transformer Models

Data Augmentation using Pre-trained Transformer Model (BERT, GPT2, etc) | Research Paper Walkthrough

Semantic Data Augmentation for End-to-End Mandarin Speech Recognition - (3 minutes introduction)...

Vorlesung 7.2 - Convolutional Neural Networks 2: Training, Data Augmentation/Dopout

Easy Data Augmentation for Text Classification

Data Augmentation in Computer Vision (CV) | Handle Imbalanced CV Data | Image Augmentation

Regularization in a Neural Network | Dealing with overfitting

AAAI 2021 KDF'21 Workshop - Data Augmentation Methods for Reject Inference in Credit Risk Model...

[en] Elyes Manai - Data Augmentation in NLP

Data-Augmentation | Keras | Arguments explained step by step | Python

Training GANs with Limited Data

MSMix An Interpolation Based Text Data Augmentation Method Manifold Swap Mixup

Mixup Data augmentation with TensorFlow 2 with intergration in tf.data - Full Stack Deep Learning.

LiDA Language Independent Data Augmentation for Text Classification final year projects