The method to map unaligned videos from multiple sources to a common representation using self-supervised objectives constructed over both time and modality (i.e. vision and sound). The embedding of YouTube videos to construct a reward function that encourages an agent to imitate human gameplay
Continue ReadingUniversal Language Model Fine-tuning for Text Classification
The article puts forward ULMFiT, effective transfer learning method that can be applied to any task in NLP. With only 100 labeled examples, it matches the performance of training from scratch on 100x more data and reduces error by 18-24% on the state-of-the-art on six text classification tasks
Continue ReadingRealistic Evaluation of Deep Semi-Supervised Learning Algorithms
Semi-supervised learning (SSL) based on deep neural networks have recently proven successful on standard benchmark tasks. Baselines which do not use unlabeled data is often underreported, SSL methods differ in sensitivity to the amount of labeled and unlabeled data, and performance can degrade substantially when the unlabeled dataset contains out-of- distribution examples. To help guide SSL research towards real-world applicability, we make our unified reimplemention and evaluation platform publicly available.
Continue Reading