← Search

Gian Paolo Leonardi

2 accepted papers

2024

A two-scale Complexity Measure for Deep Learning Models

NeurIPS 2024poster

We introduce a novel capacity measure 2sED for statistical models based on the effective dimension. The new quantity provably bounds the generalization error under mild assumptions on the model. Furthermore, simulations on standard data sets and popular model architectures show that 2sED correlates…

Cited by 1SourcePDFScholar
2022

Training Quantised Neural Networks With STE Variants: The Additive Noise Annealing Algorithm

CVPR 2022poster

Training quantised neural networks (QNNs) is a non-differentiable optimisation problem since weights and features are output by piecewise constant functions. The standard solution is to apply the straight-through estimator (STE), using different functions during the inference and gradient computatio…

Cited by 7PDFScholar