Tag
The paper presents an analysis of the environmental break-even point for ML-based data compression, estimating carbon-equivalent costs for training and inference against savings from reduced disk storage.