arXiv subjects

Wei

Publications and source records attributed to Wei.

4 recordsLinked to original sources

arXiv2025

Estimating Parameters of Structural Models Using Neural Networks

We study an alternative use of machine learning. We train neural nets to provide the parameter estimate of a given (structural) econometric model, for example, discrete choice or consumer search. Training examples consist of datasets generated by the econometric model under a range of parameter values. The neural net takes the moments of a dataset as input and tries to recognize the parameter value underlying that dataset. Besides the point estimate, the neural net can also output statistical accuracy. This neural net estimator (NNE) tends to limited-information Bayesian posterior as the number of training datasets increases. We apply NNE to a consumer search model. It gives more accurate estimates at lighter computational costs than the prevailing approach. NNE is also robust to redundant moment inputs. In general, NNE offers the most benefits in applications where other estimation approaches require very heavy simulation costs. We provide code at: https://nnehome.github.io.

econ.EM↗

arXiv2024

Inverse Design of Nonlinear Mechanics of Bio-inspired Materials Through Interface Engineering and Bayesian Optimization

In many biological materials such as nacre and bone, the material structure consists of hard grains and soft interfaces, with the interfaces playing a significant role in the material's mechanical behavior. This type of structures has been utilized in the design of various bio-inspired composite materials. Such applications often require the materials to exhibit a specified nonlinear stress-strain relationship. A key challenge lies in identifying appropriate interface properties from an infinite search space to achieve a given target stress-strain curve. This study introduces a Bayesian optimization (BO) framework specifically tailored for the inverse design of interfaces in bio-inspired composites. As a notable advantage, this method is capable of expanding the design space, allowing the discovery of optimal solutions even when the target curve deviates significantly from the initial dataset. Furthermore, our results show that BO can identify distinct interface designs that produce similar target stress-strain responses, yet differ in their deformation and failure mechanisms. These findings highlight the potential of the proposed BO framework to address a wide range of inverse design challenges in nonlinear mechanics problems.

cond-mat.mtrl-sci↗

arXiv2024

Physics-Informed Geometric Operators to Support Surrogate, Dimension Reduction and Generative Models for Engineering Design

In this work, we propose a set of physics-informed geometric operators (GOs) to enrich the geometric data provided for training surrogate/discriminative models, dimension reduction, and generative models, typically employed for performance prediction, dimension reduction, and creating data-driven parameterisations, respectively. However, as both the input and output streams of these models consist of low-level shape representations, they often fail to capture shape characteristics essential for performance analyses. Therefore, the proposed GOs exploit the differential and integral properties of shapes--accessed through Fourier descriptors, curvature integrals, geometric moments, and their invariants--to infuse high-level intrinsic geometric information and physics into the feature vector used for training, even when employing simple model architectures or low-level parametric descriptions. We showed that for surrogate modelling, along with the inclusion of the notion of physics, GOs enact regularisation to reduce over-fitting and enhance generalisation to new, unseen designs. Furthermore, through extensive experimentation, we demonstrate that for dimension reduction and generative models, incorporating the proposed GOs enriches the training data with compact global and local geometric features. This significantly enhances the quality of the resulting latent space, thereby facilitating the generation of valid and diverse designs. Lastly, we also show that GOs can enable learning parametric sensitivities to a great extent. Consequently, these enhancements accelerate the convergence rate of shape optimisers towards optimal solutions.

cs.LG↗

arXiv2023

Controllable One-Shot Face Video Synthesis With Semantic Aware Prior

The one-shot talking-head synthesis task aims to animate a source image to another pose and expression, which is dictated by a driving frame. Recent methods rely on warping the appearance feature extracted from the source, by using motion fields estimated from the sparse keypoints, that are learned in an unsupervised manner. Due to their lightweight formulation, they are suitable for video conferencing with reduced bandwidth. However, based on our study, current methods suffer from two major limitations: 1) unsatisfactory generation quality in the case of large head poses and the existence of observable pose misalignment between the source and the first frame in driving videos. 2) fail to capture fine yet critical face motion details due to the lack of semantic understanding and appropriate face geometry regularization. To address these shortcomings, we propose a novel method that leverages the rich face prior information, the proposed model can generate face videos with improved semantic consistency (improve baseline by $7\%$ in average keypoint distance) and expression-preserving (outperform baseline by $15 \%$ in average emotion embedding distance) under equivalent bandwidth. Additionally, incorporating such prior information provides us with a convenient interface to achieve highly controllable generation in terms of both pose and expression.

cs.CV↗