Abstract:Due to the difficulty of collecting paired Low-Resolution (LR) and High-Resolution (HR) images, the recent research on single image Super-Resolution (SR) has often been criticized for the data bottleneck of the synthetic image degradation between LRs and HRs. Recently, the emergence of real-world SR datasets, e.g., RealSR and DRealSR, promotes the exploration of Real-World image Super-Resolution (RWSR). RWSR exposes a more practical image degradation, which greatly challenges the learning capacity of deep neural networks to reconstruct high-quality images from low-quality images collected in realistic scenarios. In this paper, we explore Taylor series approximation in prevalent deep neural networks for image reconstruction, and propose a very general Taylor architecture to derive Taylor Neural Networks (TNNs) in a principled manner. Our TNN builds Taylor Modules with Taylor Skip Connections (TSCs) to approximate the feature projection functions, following the spirit of Taylor Series. TSCs introduce the input connected directly with each layer at different layers, to sequentially produces different high-order Taylor maps to attend more image details, and then aggregate the different high-order information from different layers. Only via simple skip connections, TNN is compatible with various existing neural networks to effectively learn high-order components of the input image with little increase of parameters. Furthermore, we have conducted extensive experiments to evaluate our TNNs in different backbones on two RWSR benchmarks, which achieve a superior performance in comparison with existing baseline methods.

Mind the Remainder: Taylor’s Theorem View on Recurrent Neural Networks

Mind the Remainder: Taylor's Theorem View on Recurrent Neural Networks

Residual Recurrent Neural Networks for Learning Sequential Representations.

Towards Interpreting Recurrent Neural Networks Through Probabilistic Abstraction

Taylor Neural Network for Real-World Image Super-Resolution

Improving RNN transducer with normalized jointer network

Character Recognition from Trajectory by Recurrent Spiking Neural Networks

Reversible Recurrent Neural Networks

Taylor saves for later: disentanglement for video prediction using Taylor representation

Learning Compact Recurrent Neural Networks with Block-Term Tensor Decomposition

NRTR: A No-Recurrence Sequence-to-Sequence Model For Scene Text Recognition

Alleviate Exposure Bias in Sequence Prediction \\ with Recurrent Neural Networks

Alleviate Exposure Bias in Sequence Prediction \\ with Recurrent Neural Networks

Developing RNN-T Models Surpassing High-Performance Hybrid Models with Customization Capability.

TTH-RNN: Tensor-Train Hierarchical Recurrent Neural Network for Video Summarization

Compressing Recurrent Neural Networks with Tensor Ring for Action Recognition

Incorporating Taylor Series and Recursive Structure in Neural Networks for Time Series Prediction

An Improved Time Feedforward Connections Recurrent Neural Networks

Understanding Recurrent Neural Networks Using Nonequilibrium Response Theory

Non-local Recurrent Neural Memory for Supervised Sequence Modeling

Exploiting Symmetric Temporally Sparse BPTT for Efficient RNN Training