XNet is a single-layer neural network architecture that leverages Cauchy integral-based activation functions for high-order function approximation. Through theoretical analysis, we show that the Cauchy activation functions used in XNet can achieve arbitrary-order polynomial convergence, fundamentally outperforming traditional MLPs and Kolmogorov-Arnold Networks (KANs) that rely on increased depth or B-spline activations. Our extensive experiments on function approximation, PDE solving, and reinforcement learning demonstrate XNet's superior performance - reducing approximation error by up to 50000 times and accelerating training by up to 10 times compared to existing approaches. These results establish XNet as a highly efficient architecture for both scientific computing and AI applications.

We initially proposed a novel method for constructing real networks from the complex domain using the Cauchy integral formula in Li et al. (2024); Zhang et al. (2024), utilizing Cauchy kernels as basis functions. This work comprehensively compares these networks with KANs, which use B-spline as basis functions in Liu et al. (2024), and MLPs to highlight our significant improvements. Multi-layer perceptrons (MLPs) (Haykin (1994); Cybenko (1989); Hornik et al. (1989)), recognized as fundamental building blocks in deep learning, have their limitations despite their wide use, particularly in its accuracy, and large number of parameters needed in structures such as in transformers (Vaswani et al. (2017)), and lack interpretability without post-analysis tools (Cunningham et al. (2023)). The Kolmogorov-Arnold Networks (KANs) were introduced as a potential alternative, drawing on the Kolmogorov-Arnold representation theorem (Kolmogorov (1956); Braun & Griebel (2009)), and demonstrate their efficiency and accuracy in computational tasks, especially in solving PDEs and function approximation (Sprecher & Draghici (2002); Köppen (2002); Lin & Unbehauen (1993); Lai & Shen (2021); Leni et al. (2013); Fakhoury et al. (2022)). In the swiftly advancing domain of deep learning, the continuous search for novel neural network designs that deliver superior accuracy and efficiency is pivotal. While traditional activation functions such as the Rectified Linear Unit (ReLU) (Nair & Hinton (2010)) have been widely adopted due to their straightforwardness and efficacy in diverse applications, their shortcomings become evident as the complexity of challenges escalates. This is particularly true in areas that demand meticulous data fitting and the solutions of intricate partial differential equations (PDEs).

XNet introduces a wavelet-based X-shaped unified architecture for fully- and semi-supervised biomedical segmentation. So far, however, XNet still faces the limitations, including performance degradation when images lack high-frequency (HF) information, underutilization of raw images and insufficient fusion. To address these issues, we propose XNet v2, a low- and high-frequency complementary model. XNet v2 performs wavelet-based image-level complementary fusion, using fusion results along with raw images inputs three different sub-networks to construct consistency loss. Furthermore, we introduce a feature-level fusion module to enhance the transfer of low-frequency (LF) information and HF information. XNet v2 achieves state-of-the-art in semi-supervised segmentation while maintaining competitve results in fully-supervised learning. More importantly, XNet v2 excels in scenarios where XNet fails. Compared to XNet, XNet v2 exhibits fewer limitations, better results and greater universality. Extensive experiments on three 2D and two 3D datasets demonstrate the effectiveness of XNet v2. Code is available at https://github.com/Yanfeng-Zhou/XNetv2 .

XPENG is preparing to launch its own robotaxi fleet, the world's first fully-electric flying car, and an AI-powered robot. The luxury vehicle maker said that its latest G9 SUV became China's first mass-produced commercial vehicle to pass the government-led Autonomous Driving Closed-field Test. "Obtaining the road test permit by our mass-produced commercial vehicles – with no retrofit – is a major achievement," said Dr Wu on XPENG's fourth annual Tech Day this week. "Our platform-based robotaxi development aims to generate significant cost benefits, and ensure product quality, safety, and user experience." The G9 uses XNGP, an advanced driver assistance platform developed by XPENG that supports driving scenarios from highways to complex urban roads.

Recent GAN-based architectures have been able to deliver impressive performance on the general task of image-to-image translation. In particular, it was shown that a wide variety of image translation operators may be learned from two image sets, containing images from two different domains, without establishing an explicit pairing between the images. This was made possible by introducing clever regularizers to overcome the under-constrained nature of the unpaired translation problem. In this work, we introduce a novel architecture for unpaired image translation, and explore several new regularizers enabled by it. Specifically, our architecture comprises a pair of GANs, as well as a pair of translators between their respective latent spaces. These cross-translators enable us to impose several regularizing constraints on the learnt image translation operator, collectively referred to as latent cross-consistency. Our results show that our proposed architecture and latent cross-consistency constraints are able to outperform the existing state-of-the-art on a wide variety of image translation tasks.

X-Ray image enhancement, along with many other medical image processing applications, requires the segmentation of images into bone, soft tissue, and open beam regions. We apply a machine learning approach to this problem, presenting an end-to-end solution which results in robust and efficient inference. Since medical institutions frequently do not have the resources to process and label the large quantity of X-Ray images usually needed for neural network training, we design an end-to-end solution for small datasets, while achieving state-of-the-art results. Our implementation produces an overall accuracy of 92%, F1 score of 0.92, and an AUC of 0.98, surpassing classical image processing techniques, such as clustering and entropy based methods, while improving upon the output of existing neural networks used for segmentation in non-medical contexts. The code used for this project is available online.