Pith. sign in

Practical Gauss-Newton Optimisation for Deep Learning

1 Pith paper cite this work, alongside 35 external citations. Polarity classification is still indexing.

1 Pith paper citing it
35 external citations · Pith
abstract

We present an efficient block-diagonal ap- proximation to the Gauss-Newton matrix for feedforward neural networks. Our result- ing algorithm is competitive against state- of-the-art first order optimisation methods, with sometimes significant improvement in optimisation performance. Unlike first-order methods, for which hyperparameter tuning of the optimisation parameters is often a labo- rious process, our approach can provide good performance even when used with default set- tings. A side result of our work is that for piecewise linear transfer functions, the net- work objective function can have no differ- entiable local maxima, which may partially explain why such transfer functions facilitate effective optimisation.

citation-role summary

background 1

citation-polarity summary

fields

stat.ME 1

years

2026 1

verdicts

CONDITIONAL 1

roles

background 1

polarities

unclear 1

representative citing papers

Parameter identification for predator-prey system with sparse data

stat.ME · 2026-08-05 · conditional · novelty 5.0

A parameter-estimation framework for Lotka-Volterra models with sparse data, combining Natural Gradient Ascent, non-dimensionalization, and an adaptive ODE solver, claims fewer iterations and more stability than standard methods.

citing papers explorer

Showing 1 of 1 citing paper.

  • Parameter identification for predator-prey system with sparse data stat.ME · 2026-08-05 · conditional · none · ref 2 · internal anchor

    A parameter-estimation framework for Lotka-Volterra models with sparse data, combining Natural Gradient Ascent, non-dimensionalization, and an adaptive ODE solver, claims fewer iterations and more stability than standard methods.