Normalizing Flows (NFs) are a class of generative models distinguished by a mathematically invertible architecture, where the forward pass transforms data into a latent space for density estimation, and the reverse pass generates new samples from this space. This characteristic creates an intrinsic synergy between representation learning and data generation. However, the generative quality of standard NFs is limited by poor semantic representations from log-likelihood optimization. To remedy this, we propose a novel alignment strategy that creatively leverages the invertibility of NFs: instead of regularizing the forward pass, we align the intermediate features of the generative (reverse) pass with representations from a powerful vision foundation model, demonstrating superior effectiveness over naive alignment. We also introduce a novel training-free, test-time optimization algorithm for classification, which provides a more intrinsic evaluation of the NF's embedded semantic knowledge. Comprehensive experiments demonstrate that our approach accelerates the training of NFs by over 3.3x, while simultaneously delivering significant improvements in both generative quality and classification accuracy. New state-of-the-art results for NFs are established on ImageNet 64 x 64 and 256 x 256.

In medical image segmentation, domain generalization poses a significant challenge due to domain shifts caused by variations in data acquisition devices and other factors. These shifts are particularly pronounced in the most common scenario, which involves only single-source domain data due to privacy concerns. To address this, we draw inspiration from the self-supervised learning paradigm that effectively discourages overfitting to the source domain. We propose the Denoising Y-Net (DeY-Net), a novel approach incorporating an auxiliary denoising decoder into the basic U-Net architecture. The auxiliary decoder aims to perform denoising training, augmenting the domain-invariant representation that facilitates domain generalization. Furthermore, this paradigm provides the potential to utilize unlabeled data. Building upon denoising training, we propose Denoising Test Time Adaptation (DeTTA) that further: (i) adapts the model to the target domain in a sample-wise manner, and (ii) adapts to the noise-corrupted input. Extensive experiments conducted on widely-adopted liver segmentation benchmarks demonstrate significant domain generalization improvements over our baseline and state-of-the-art results compared to other methods. Code is available at https://github.com/WenRuxue/DeTTA.

With the development of location-based services (LBS), concerns on location privacy frequently arise. Location data often contains users' sensitive information, and direct release it may pose a threat to users' privacy. Differential privacy (DP), as a privacy preserving method with solid mathematical foundation, has been widely used in location data release. However, most if not all of the existing location DP mechanisms only consider static scenarios or perturb the location at single timestamp, which are vulnerable to the so-called location-dependent attacks (LDA) in continuous LBS queries. In this paper, an optimal location DP mechanism against LDA is proposed. Firstly, the necessary conditions for LDA defense are derived by combining the perturbation mechanism of location DP. Then the algorithm of safe perturbance region generation is established to dynamically calculate the perturbation range at each timestamp. Finally, we set up the optimization problem with the real-time quality loss as the optimization objective and the location DP and safe perturbance region as the optimization conditions, and realize the optimal DP mechanism for LDA by solving the optimization problem. Experiment results on real-world datasets show that our mechanism can effectively resist LDA, which also balance privacy protection and data utility well.