Collapsing of dimensionality
This work addresses the challenge of online learning and parameter estimation in machine learning, though it appears incremental as it builds on existing regularization networks.
The authors tackled the problem of learning model parameters by proposing a new approach that modifies classical regularization networks to incorporate temporal dynamics, leading to a dimensionality collapse. They reported extensive experimental exploration on artificial datasets to evaluate the model's behavior.
We analyze a new approach to Machine Learning coming from a modification of classical regularization networks by casting the process in the time dimension, leading to a sort of collapse of dimensionality in the problem of learning the model parameters. This approach allows the definition of a online learning algorithm that progressively accumulates the knowledge provided in the input trajectory. The regularization principle leads to a solution based on a dynamical system that is paired with a procedure to develop a graph structure that stores the input regularities acquired from the temporal evolution. We report an extensive experimental exploration on the behavior of the parameter of the proposed model and an evaluation on artificial dataset.