Abstract:U-Nets are a go-to, state-of-the-art neural architecture across numerous tasks for continuous signals on a square such as images and Partial Differential Equations (PDE), however their design and architecture is understudied. In this paper, we provide a framework for designing and analysing general U-Net architectures. We present theoretical results which characterise the role of the encoder and decoder in a U-Net, their high-resolution scaling limits and their conjugacy to ResNets via preconditioning. We propose Multi-ResNets, U-Nets with a simplified, wavelet-based encoder without learnable parameters. Further, we show how to design novel U-Net architectures which encode function constraints, natural bases, or the geometry of the data. In diffusion models, our framework enables us to identify that high-frequency information is dominated by noise exponentially faster, and show how U-Nets with average pooling exploit this. In our experiments, we demonstrate how Multi-ResNets achieve competitive and often superior performance compared to classical U-Nets in image segmentation, PDE surrogate modelling, and generative modelling with diffusion models. Our U-Net framework paves the way to study the theoretical properties of U-Nets and design natural, scalable neural architectures for a multitude of problems beyond the square.

The general critical analysis for continuous-time UPPAM recurrent neural networks.

The UPPAM Continuous-Time RNN Model and Its Critical Dynamics Study

Without Diagonal Nonlinear Requirements: the More General P-Critical Dynamical Analysis for UPPAM Recurrent Neural Networks

Convergence Analysis of General Neural Networks under Almost Periodic Stimuli

On the Flexible Dynamics Analysis for the Unified Discrete-Time RNNs

Towards establishing a meaningful and practical dynamics results for the unified RNN model.

A Critical Global Convergence Analysis of Recurrent Neural Networks with General Projection Mappings

New Critical Analysis on Global Convergence of Recurrent Neural Networks with Projection Mappings

Critical Dynamical Analysis for Α-Uam RNNs Without Diagonal Nonlinear Requirements.

Towards a Unified Recurrent Neural Network Theory: the Uniformly Pseudo-Projection-anti-monotone Net

On the P-critical dynamics analysis of projection recurrent neural networks

Dynamics Analysis for Generic Projection Continuous-Time RNNs with Bounded Matrices

New Approach for Robust Stability Analysis of Delayed Recurrent Neural Networks

LMI-based Approach for Global Asymptotic Stability Analysis of Continuous BAM Neural Networks

A New Neural Network Model And Its Application

Critical dynamics study on recurrent neural networks: Globally exponential stability

Continuous Attractors of Discrete-Time Recurrent Neural Networks

Convergence Theory For Unified Discrete-Time Rnns With Quasi-Symmetric Connection

A Unified Framework for U-Net Design and Analysis

Stability Analysis on a Class of Nonlinear Continuous Neural Networks

Finite-time peak-to-peak analysis for switched generalized neural networks comprised of finite-time unstable subnetworks