SemDeDup: Data-efficient learning at web-scale through semantic deduplication
This addresses data inefficiency for machine learning practitioners by enabling faster training with less data, though it is incremental as it builds on existing embedding techniques.
The authors tackled the problem of redundancy in large web-scale datasets by introducing SemDeDup, a method that removes semantic duplicates using pre-trained embeddings, resulting in up to 50% data reduction with minimal performance loss and halved training time, while improving out-of-distribution performance.
Progress in machine learning has been driven in large part by massive increases in data. However, large web-scale datasets such as LAION are largely uncurated beyond searches for exact duplicates, potentially leaving much redundancy. Here, we introduce SemDeDup, a method which leverages embeddings from pre-trained models to identify and remove semantic duplicates: data pairs which are semantically similar, but not exactly identical. Removing semantic duplicates preserves performance and speeds up learning. Analyzing a subset of LAION, we show that SemDeDup can remove 50% of the data with minimal performance loss, effectively halving training time. Moreover, performance increases out of distribution. Also, analyzing language models trained on C4, a partially curated dataset, we show that SemDeDup improves over prior approaches while providing efficiency gains. SemDeDup provides an example of how simple ways of leveraging quality embeddings can be used to make models learn faster with less data.