Search for: All records

Modeling population dynamics is a fundamental problem with broad scientific applications. Motivated by real-world applications including biosystems with diverse populations, we consider a class of population dynamics modeling with two technical challenges: (i) dynamics to learn for individual particles are heterogeneous and (ii) available data to learn from are not time-series (i.e, each individual’s state trajectory over time) but cross-sectional (i.e, the whole population’s aggregated states without individuals matched over time). To address the challenges, we introduce a novel computational framework dubbed correlational Lagrangian Schrödinger bridge (CLSB) that builds on optimal transport to “bridge" cross-sectional data distributions. In contrast to prior methods regularizing all individuals’ transport “costs” and then applying them to the population homogeneously, CLSB directly regularizes population cost allowing for population heterogeneity and potentially improving model generalizability. Specifically our contributions include (1) a novel population perspective of the transport cost and a new class of population regularizers capturing the temporal variations in multivariate relations, with the tractable formulation derived, (2) three domain-informed instantiations of population regularizers on covariance, and (3) integration of population regularizers into data-driven generative models as constrained optimization and an approximate numerical solution, with further extension to conditional generative models. Empirically, we demonstrate the superiority of CLSB in single-cell sequencing data analyses (including cell differentiation and drug-conditioned cell responses) and opinion depolarization. 

Abstract The evolutionary path of massive stars begins at helium burning. Energy production for this phase of stellar evolution is dominated by the reaction path 3$$\alpha \rightarrow ^{12}$$ $\alpha {\to }^{12}$C$$(\alpha ,\gamma )^{16}$$ ${(\alpha ,\gamma )}^{16}$O and also determines the ratio of$$^{12}$$ $_{}^{12}$C/$$^{16}$$ $_{}^{16}$O in the stellar core. This ratio then sets the evolutionary trajectory as the star evolves towards a white dwarf, neutron star or black hole. Although the reaction rate of the 3$$\alpha $$ $\alpha$process is relatively well known, since it proceeds mainly through a single narrow resonance in$$^{12}$$ $_{}^{12}$C, that of the$$^{12}$$ $_{}^{12}$C$$(\alpha ,\gamma )^{16}$$ ${(\alpha ,\gamma )}^{16}$O reaction remains uncertain since it is the result of a more difficult to pin down, slowly-varying, portion of the cross section over a strong interference region between the high-energy tails of subthreshold resonances, the low-energy tails of higher-energy broad resonances and direct capture. Experimental measurements of this cross section require herculean efforts, since even at higher energies the cross section remains small and large background sources are often present that require the use of very sensitive experimental methods. Since the$$^{12}$$ $_{}^{12}$C$$(\alpha ,\gamma )^{16}$$ ${(\alpha ,\gamma )}^{16}$O reaction has such a strong influence on many different stellar objects, it is also interesting to try to back calculate the required rate needed to match astrophysical observations. This has become increasingly tempting, as the accuracy and precision of observational data has been steadily improving. Yet, the pitfall to this approach lies in the intermediary steps of modeling, where other uncertainties needed to model a star’s internal behavior remain highly uncertain. 

This paper considers the problem of offline optimization, where the objective function is unknown except for a collection of “offline" data examples. While recent years have seen a flurry of work on applying various machine learning techniques to the offline optimization problem, the majority of these works focused on learning a surrogate of the unknown objective function and then applying existing optimization algorithms. While the idea of modeling the unknown objective function is intuitive and appealing, from the learning point of view it also makes it very difficult to tune the objective of the learner according to the objective of optimization. Instead of learning and then optimizing the unknown objective function, in this paper we take on a less intuitive but more direct view that optimization can be thought of as a process of sampling from a generative model. To learn an effective generative model from the offline data examples, we consider the standard technique of “re-weighting", and our main technical contribution is a probably approximately correct (PAC) lower bound on the natural optimization objective, which allows us to jointly learn a weight function and a score-based generative model from a surrogate loss function. The robustly competitive performance of the proposed approach is demonstrated via empirical studies using the standard offline optimization benchmarks. 

Aims.The goal of this project is to construct an estimator for the masses of supermassive black holes in active galactic nuclei (AGNs) based on the broad Hαemission line. Methods.We made use of published reverberation mapping data. We remeasured all Hαtime lags from the original data as we find that reverberation measurements are often improved by detrending the light curves. Results.We produced mass estimators that require only the Hαluminosity and the width of the Hαemission line as characterized by either the full width at half maximum or the line dispersion. Conclusions.It is possible, on the basis of a single spectrum covering the Hαemission line, to estimate the mass of the central supermassive black hole in AGNs with all three parameters believed to affect mass measurement – luminosity, line width, and Eddington ratio – taken into account. The typical formal accuracy in such estimates is of order 0.2–0.3 dex relative to the reverberation-based masses. 

Generating 3D graphs of symmetry-group equivariance is of intriguing potential in broad applications from machine vision to molecular discovery. Emerging approaches adopt diffusion generative models (DGMs) with proper re-engineering to capture 3D graph distributions. In this paper, we raise an orthogonal and fundamental question of in what (latent) space we should diffuse 3D graphs. ❶ We motivate the study with theoretical analysis showing that the performance bound of 3D graph diffusion can be improved in a latent space versus the original space, provided that the latent space is of (i) low dimensionality yet (ii) high quality (i.e., low reconstruction error) and DGMs have (iii) symmetry preservation as an inductive bias. ❷ Guided by the theoretical guidelines, we propose to perform 3D graph diffusion in a low-dimensional latent space, which is learned through cascaded 2D–3D graph autoencoders for low-error reconstruction and symmetry-group invariance. The overall pipeline is dubbed latent 3D graph diffusion. ❸ Motivated by applications in molecular discovery, we further extend latent 3D graph diffusion to conditional generation given SE(3)-invariant attributes or equivariant 3D objects. ❹ We also demonstrate empirically that out-of-distribution conditional generation can be further improved by regularizing the latent space via graph self-supervised learning. We validate through comprehensive experiments that our method generates 3D molecules of higher validity / drug-likeliness and comparable or better conformations / energetics, while being an order of magnitude faster in training. Codes are released at https://github.com/Shen-Lab/LDM-3DG.