Unraveling the Enigma of Double Descent: An In-depth Analysis through the Lens of Learned Feature Space
This addresses a fundamental puzzle in deep learning theory for researchers, though it appears incremental as it builds on prior explanations.
The paper tackles the double descent phenomenon in machine learning by analyzing learned feature spaces, finding that it arises in imperfect models trained with noisy data, with the model first learning noise until interpolation and then using over-parameterization to separate information from noise.
Double descent presents a counter-intuitive aspect within the machine learning domain, and researchers have observed its manifestation in various models and tasks. While some theoretical explanations have been proposed for this phenomenon in specific contexts, an accepted theory to account for its occurrence in deep learning remains yet to be established. In this study, we revisit the phenomenon of double descent and demonstrate that its occurrence is strongly influenced by the presence of noisy data. Through conducting a comprehensive analysis of the feature space of learned representations, we unveil that double descent arises in imperfect models trained with noisy data. We argue that double descent is a consequence of the model first learning the noisy data until interpolation and then adding implicit regularization via over-parameterization acquiring therefore capability to separate the information from the noise.