More Like this

1. Optimal Design of Bacterial Carpets for Fluid Pumping

   https://doi.org/10.3390/fluids7010025  

   Rostami, Minghao W.; Liu, Weifan; Buchmann, Amy; Strawbridge, Eva; Zhao, Longhua (January 2022, Fluids)

   In this work, we outline a methodology for determining optimal helical flagella placement and phase shift that maximize fluid pumping through a rectangular flow meter above a simulated bacterial carpet. This method uses a Genetic Algorithm (GA) combined with a gradient-based method, the Broyden-Fletcher-Goldfarb-Shanno (BFGS) algorithm, to solve the optimization problem and the Method of Regularized Stokeslets (MRS) to simulate the fluid flow. This method is able to produce placements and phase shifts for small carpets and could be adapted for implementation in larger carpets and various fluid tasks. Our results show that given identical helices, optimal pumping configurations are influenced by the size of the flow meter. We also show that intuitive designs, such as uniform placement, do not always lead to a high-performance carpet. 

   more » « less
2. Hierarchical Shrinkage: Improving the Accuracy and Interpretability of Tree-Based Methods

   Agarwal, Abhineet; Tan, Yan Shuo; Ronen, Omer; Singh, Chandan; Yu, Bin (June 2022, Proceedings of Machine Learning Research)

   Tree-based models such as decision trees and random forests (RF) are a cornerstone of modern machine-learning practice. To mitigate overfitting, trees are typically regularized by a variety of techniques that modify their structure (e.g. pruning). We introduce Hierarchical Shrinkage (HS), a post-hoc algorithm that does not modify the tree structure, and instead regularizes the tree by shrinking the prediction over each node towards the sample means of its ancestors. The amount of shrinkage is controlled by a single regularization parameter and the number of data points in each ancestor. Since HS is a post-hoc method, it is extremely fast, compatible with any tree growing algorithm, and can be used synergistically with other regularization techniques. Extensive experiments over a wide variety of real world datasets show that HS substantially increases the predictive performance of decision trees, even when used in conjunction with other regularization techniques. Moreover, we find that applying HS to each tree in an RF often improves accuracy, as well as its interpretability by simplifying and stabilizing its decision boundaries and SHAP values. We further explain the success of HS in improving prediction performance by showing its equivalence to ridge regression on a (supervised) basis constructed of decision stumps associated with the internal nodes of a tree. All code and models are released in a full fledged package available on Github. 

   more » « less
3. Stochastic Variance-Reduced Cubic Regularized Newton Method

   Zhou, Dongruo; Xu, Pan; Gu, Quanquan (July 2018, International Conference on Machine Learning)

   We propose a stochastic variance-reduced cubic regularized Newton method (SVRC) for non-convex optimization. At the core of our algorithm is a novel semi-stochastic gradient along with a semi-stochastic Hessian, which are specifically designed for cubic regularization method. We show that our algorithm is guaranteed to converge to an $$(\epsilon,\sqrt{\epsilon})$$-approximate local minimum within $$\tilde{O}(n^{4/5}/\epsilon^{3/2})$$ second-order oracle calls, which outperforms the state-of-the-art cubic regularization algorithms including subsampled cubic regularization. Our work also sheds light on the application of variance reduction technique to high-order non-convex optimization methods. Thorough experiments on various non-convex optimization problems support our theory. 

   more » « less
4. Remarks on Regularized Stokeslets in Slender Body Theory

   https://doi.org/10.3390/fluids6080283  

   Ohm, Laurel (August 2021, Fluids)

   We remark on the use of regularized Stokeslets in the slender body theory (SBT) approximation of Stokes flow about a thin fiber of radius ϵ>0. Denoting the regularization parameter by δ, we consider regularized SBT based on the most common regularized Stokeslet plus a regularized doublet correction. Given sufficiently smooth force data along the filament, we derive L∞ bounds for the difference between regularized SBT and its classical counterpart in terms of δ, ϵ, and the force data. We show that the regularized and classical expressions for the velocity of the filament itself differ by a term proportional to log(δ/ϵ); in particular, δ=ϵ is necessary to avoid an O(1) discrepancy between the theories. However, the flow at the surface of the fiber differs by an expression proportional to log(1+δ2/ϵ2), and any choice of δ∝ϵ will result in an O(1) discrepancy as ϵ→0. Consequently, the flow around a slender fiber due to regularized SBT does not converge to the solution of the well-posed slender body PDE which classical SBT is known to approximate. Numerics verify this O(1) discrepancy but also indicate that the difference may have little impact in practice. 

   more » « less
5. Finite difference method in prolate spheroidal coordinates for freely suspended spheroidal particles in linear flows of viscous and viscoelastic fluids

   https://doi.org/10.1016/j.jcp.2023.112559  

   Sharma, Arjun; Koch, Donald L (December 2023, Journal of Computational Physics)

   A finite difference scheme is used to develop a numerical method to solve the flow of an unbounded viscoelastic fluid with zero to moderate inertia around a prolate spheroidal particle. The equations are written in prolate spheroidal coordinates, and the shape of the particle is exactly resolved as one of the coordinate surfaces representing the inner boundary of the computational domain. As the prolate spheroidal grid is naturally clustered near the particle surface, good resolution is obtained in the regions where the gradients of relevant flow variables are most significant. This coordinate system also allows large domain sizes with a reasonable number of mesh points to simulate unbounded fluid around a particle. Changing the aspect ratio of the inner computational boundary enables simulations of different particle shapes ranging from a sphere to a slender fiber. Numerical studies of the latter particle shape allow testing of slender body theories. The mass and momentum equations are solved with a Schur complement approach allowing us to solve the zero inertia case necessary to isolate the viscoelastic effects. The singularities associated with the coordinate system are overcome using L’Hopital’s rule. A straightforward imposition of conditions representing a time-varying combination of linear flows on the outer boundary allows us to study various flows with the same computational domain geometry. For the special but important case of zero fluid and particle inertia we obtain a novel formulation that satisfies the force- and torque-free constraint in an iteration-free manner. The numerical method is demonstrated for various flows of Newtonian and viscoelastic fluids around spheres and spheroids (including those with large aspect ratio). Good agreement is demonstrated with existing theoretical and numerical results. 

   more » « less