[D] Negative examples are still useful in self-supervised learning even after the BYOL, and they are directly trainable end-to-end with a backbone.

• AdCo

  2 161 0.9 Python

  AdCo: Adversarial Contrast for Efficient Learning of Unsupervised Representations from Self-Trained Negative Adversaries

  

The paper showed that with only 8196 negatives, the AdCo can achieve better performance than the SOTA self-supervised methods (MoCo V2, SimCLR, AdCo and SWAV) with fewer epochs, thus making the AdCo a very efficient self-supervised learning algorithm to pretrain a backbone. The source code has been released at https://github.com/maple-research-lab/AdCo.

• InfluxDB

  www.influxdata.com featured

  Power Real-Time Data Analytics at Scale. Get real-time insights from all types of time series data with InfluxDB. Ingest, query, and analyze billions of data points in real-time with unbounded cardinality.

  

Suggest a related project