Optimization in the Bures-Wasserstein space has been gaining popularity in the machine learning community since it draws connections between variational inference and Wasserstein gradient flows. The variational inference objective function of Kullback–Leibler divergence can be written as the sum of the negative entropy and the potential energy, making forward-backward Euler the method of choice. Notably, the backward step admits a closed-form solution in this case, facilitating the practicality of the scheme. However, the forward step is not exact since the Bures-Wasserstein gradient of the potential energy involves "intractable" expectations. Recent approaches propose using the Monte Carlo method -- in practice a single-sample estimator -- to approximate these terms, resulting in high variance and poor performance. We propose a novel variance-reduced estimator based on the principle of control variates. We theoretically show that this estimator has a smaller variance than the Monte-Carlo estimator in scenarios of interest. We also prove that variance reduction helps improve the optimization bounds of the current analysis. We demonstrate that the proposed estimator gains order-of-magnitude improvements over the previous Bures-Wasserstein methods.

Density ratio estimation in high dimensions can be reframed as integrating a certain quantity, the time score, over probability paths which interpolate between the two densities. In practice, the time score has to be estimated based on samples from the two densities. However, existing methods for this problem remain computationally expensive and can yield inaccurate estimates. Inspired by recent advances in generative modeling, we introduce a novel framework for time score estimation, based on a conditioning variable. Choosing the conditioning variable judiciously enables a closed-form objective function. We demonstrate that, compared to previous approaches, our approach results in faster learning of the time score and competitive or better estimation accuracies of the density ratio on challenging tasks. Furthermore, we establish theoretical guarantees on the error of the estimated density ratio.

Neural models learn representations of high-dimensional data on low-dimensional manifolds. Multiple factors, including stochasticities in the training process, model architectures, and additional inductive biases, may induce different representations, even when learning the same task on the same data. However, it has recently been shown that when a latent structure is shared between distinct latent spaces, relative distances between representations can be preserved, up to distortions. Building on this idea, we demonstrate that exploiting the differential-geometric structure of latent spaces of neural models, it is possible to capture *precisely* the transformations between representational spaces trained on similar data distributions. Specifically, we assume that distinct neural models parametrize approximately the same underlying manifold, and introduce a representation based on the *pullback metric* that captures the intrinsic structure of the latent space, while scaling efficiently to large models. We validate experimentally our method on model stitching and retrieval tasks, covering autoencoders and vision foundation discriminative models, across diverse architectures, datasets, pretraining schemes and modalities. Code is available at the following [link](https://github.com/marc0git/RelativeGeodesics).

Traditional Markov Chain Monte Carlo sampling methods often struggle with sharp curvatures, intricate geometries, and multimodal distributions. Slice sampling can resolve local exploration inefficiency issues, and Riemannian geometries help with sharp curvatures. Recent extensions enable slice sampling on Riemannian manifolds, but they are restricted to cases where geodesics are available in a closed form. We propose a method that generalizes Hit-and-Run slice sampling to more general geometries tailored to the target distribution, by approximating geodesics as solutions to differential equations. Our approach enables the exploration of the regions with strong curvature and rapid transitions between modes in multimodal distributions. We demonstrate the advantages of the approach over challenging sampling problems.

Optimization-based controller tuning is challenging because it requires formulating optimization problems explicitly as functions of controller parameters. Safe learning algorithms overcome the challenge by creating surrogate models from measured data. To ensure safety, such data-driven algorithms often rely on exhaustive grid search, which is computationally inefficient. In this paper, we propose a novel approach to safe learning by formulating a series of optimization problems instead of a grid search. We also develop a method for initializing the optimization problems to guarantee feasibility while using numerical solvers. The performance of the new method is first validated in a simulated precision motion system, demonstrating improved computational efficiency, and illustrating the role of exploiting numerical solvers to reach the desired precision. Experimental validation on an industrial-grade precision motion system confirms that the proposed algorithm achieves 30% better tracking at sub-micrometer precision as a state-of-the-art safe learning algorithm, improves the default auto-tuning solution, and reduces the computational cost seven times compared to learning algorithms based on exhaustive search.

Laplace’s method approximates a target density with a Gaussian distribution at its mode. It is computationally efficient and asymptotically exact for Bayesian inference due to the Bernstein-von Mises theorem, but for complex targets and finite-data posteriors it is often too crude an approximation. A recent generalization of the Laplace Approximation transforms the Gaussian approximation according to a chosen Riemannian geometry providing a richer approximation family, while still retaining computational efficiency. However, as shown here, its properties depend heavily on the chosen metric, indeed the metric adopted in previous work results in approximations that are overly narrow as well as being biased even at the limit of infinite data. We correct this shortcoming by developing the approximation family further, deriving two alternative variants that are exact at the limit of infinite data, extending the theoretical analysis of the method, and demonstrating practical improvements in a range of experiments.

We study a class of optimization problems in the Wasserstein space (the space of probability measures) where the objective function is nonconvex along generalized geodesics. Specifically, the objective exhibits some difference-of-convex structure along these geodesics. The setting also encompasses sampling problems where the logarithm of the target distribution is difference-of-convex. We derive multiple convergence insights for a novel semi Forward-Backward Euler scheme under several nonconvex (and possibly nonsmooth) regimes. Notably, the semi Forward-Backward Euler is just a slight modification of the Forward-Backward Euler whose convergence is---to our knowledge---still unknown in our very general non-geodesically-convex setting.

Facial recognition has become the surge on mobile authentication scenarios and makes up a huge market share for various apps, such as MasterCard, Google Wallet, and AliPay. However, existing solutions suffer from various impersonation attacks, including photo-spoofing attack, video-replay attack, and 3D facial mask attack. State-of-the-art countermeasures either require additional user intervention or introduce specialized high-end sensors. Even introducing these extra efforts, these approaches still hardly defend the latest 3D facial mask attacks, which gradually become accessible due to the prevalence of low-cost 3D printing. In this paper, we propose an anti-spoofing facial recognition system, PassFace, which verifies the smartphone for authentication as the second factor merely using raw facial videos without any user intervention, to defeat impersonation attacks. In particular, when receiving a user’s selfie video, PassFace identifies the user’s face from the video, and meanwhile extracts the highly unique and physically irreproducible camera fingerprint, i.e., Photo Response Non-Uniformity (PRNU), built in the smartphone from key frames of the video. After that, the system compares the Peak to Correlation Energy (PCE) calculated by the estimated PRNU and the reference profile with a threshold for authentication. Experiment results demonstrate PassFace can achieve satisfactory performance in authentication and attack resistance.

In this demonstration, we show that audio signals emitted by a smartphone speaker can be captured by the accelerometer on the same smartphone, and accelerometers on recently released smartphones can cover most of the fundamental frequency band of adult speech. Based on these pivotal observations, we present AccelEve, a new side channel attack that allows smartphone applications to eavesdrop on the smartphone speaker without the requirement of sensitive system permissions. Through analyzing the accelerometer measurements of a smartphone, AccelEve is able to: 1) recognize the speech information (text) carried by the acceleration signal; 2) reconstruct the audio signal played by the smartphone speaker. This demo will present experimental validations for our observations and the proposed system.