Search Results for "bootstrap"

Practical implementation of an astoundingly simple method for self-supervised learning that achieves a new state-of-the-art (surpassing SimCLR) without contrastive learning and having to designate negative pairs. This repository offers a module that one can easily wrap any image-based neural network (residual network, discriminator, policy network) to immediately start benefitting from unlabelled image data. There is now new evidence that batch normalization is key to making this technique...

The caret (Classification And Regression Training) R package streamlines the process of building predictive machine learning models. It provides uniform interfaces for model training, tuning, evaluation, preprocessing, and variable importance. With support for over 200 models, caret is foundational for R workflows in modeling and machine learning.

sense2vec (Trask et. al, 2015) is a nice twist on word2vec that lets you learn more interesting and detailed word vectors. This library is a simple Python implementation for loading, querying and training sense2vec models. For more details, check out our blog post. To explore the semantic similarities across all Reddit comments of 2015 and 2019, see the interactive demo.