-
Expert-guided Bayesian Optimisation for Human-in-the-loop Experimental Design of Known Systems
Authors:
Tom Savage,
Ehecatl Antonio del Rio Chanona
Abstract:
Domain experts often possess valuable physical insights that are overlooked in fully automated decision-making processes such as Bayesian optimisation. In this article we apply high-throughput (batch) Bayesian optimisation alongside anthropological decision theory to enable domain experts to influence the selection of optimal experiments. Our methodology exploits the hypothesis that humans are bet…
▽ More
Domain experts often possess valuable physical insights that are overlooked in fully automated decision-making processes such as Bayesian optimisation. In this article we apply high-throughput (batch) Bayesian optimisation alongside anthropological decision theory to enable domain experts to influence the selection of optimal experiments. Our methodology exploits the hypothesis that humans are better at making discrete choices than continuous ones and enables experts to influence critical early decisions. At each iteration we solve an augmented multi-objective optimisation problem across a number of alternate solutions, maximising both the sum of their utility function values and the determinant of their covariance matrix, equivalent to their total variability. By taking the solution at the knee point of the Pareto front, we return a set of alternate solutions at each iteration that have both high utility values and are reasonably distinct, from which the expert selects one for evaluation. We demonstrate that even in the case of an uninformed practitioner, our algorithm recovers the regret of standard Bayesian optimisation.
△ Less
Submitted 5 December, 2023;
originally announced December 2023.
-
Multi-Fidelity Data-Driven Design and Analysis of Reactor and Tube Simulations
Authors:
Tom Savage,
Nausheen Basha,
Jonathan McDonough,
Omar K Matar,
Ehecatl Antonio del Rio Chanona
Abstract:
The development of new manufacturing techniques such as 3D printing have enabled the creation of previously infeasible chemical reactor designs. Systematically optimizing the highly parameterized geometries involved in these new classes of reactor is vital to ensure enhanced mixing characteristics and feasible manufacturability. Here we present a framework to rapidly solve this nonlinear, computat…
▽ More
The development of new manufacturing techniques such as 3D printing have enabled the creation of previously infeasible chemical reactor designs. Systematically optimizing the highly parameterized geometries involved in these new classes of reactor is vital to ensure enhanced mixing characteristics and feasible manufacturability. Here we present a framework to rapidly solve this nonlinear, computationally expensive, and derivative-free problem, enabling the fast prototype of novel reactor parameterizations. We take advantage of Gaussian processes to adaptively learn a multi-fidelity model of reactor simulations across a number of different continuous mesh fidelities. The search space of reactor geometries is explored through an amalgam of different, potentially lower, fidelity simulations which are chosen for evaluation based on weighted acquisition function, trading off information gain with cost of simulation. Within our framework we derive a novel criteria for monitoring the progress and dictating the termination of multi-fidelity Bayesian optimization, ensuring a high fidelity solution is returned before experimental budget is exhausted. The class of reactor we investigate are helical-tube reactors under pulsed-flow conditions, which have demonstrated outstanding mixing characteristics, have the potential to be highly parameterized, and are easily manufactured using 3D printing. To validate our results, we 3D print and experimentally validate the optimal reactor geometry, confirming its mixing performance. In doing so we demonstrate our design framework to be extensible to a broad variety of expensive simulation-based optimization problems, supporting the design of the next generation of highly parameterized chemical reactors.
△ Less
Submitted 7 July, 2023; v1 submitted 1 May, 2023;
originally announced May 2023.
-
The "Hot Spots" Conjecture on the Vicsek Set
Authors:
Marius Ionescu,
Thomas L. Savage
Abstract:
We prove the Hot Spot conjecture on the Vicsek set. Specifically, we show that every eigenfunction of the second smallest eigenvalue of the Neumann Laplacian on the Vicsek set attains its maximum and minimum on the boundary.
We prove the Hot Spot conjecture on the Vicsek set. Specifically, we show that every eigenfunction of the second smallest eigenvalue of the Neumann Laplacian on the Vicsek set attains its maximum and minimum on the boundary.
△ Less
Submitted 3 January, 2019; v1 submitted 13 June, 2018;
originally announced June 2018.
-
General monotonicity, interpolation of operators, and applications
Authors:
S. M. Grigoriev,
Y. Sagher,
T. R. Savage
Abstract:
We continue the work of S. Tikhonov, E. Liflyand, B. Booton, and others, proving the equivalence of L(p,q)-norms of general monotone functions and of their Fourier transforms. The main tool in this work is the interpolation properties of cones of general monotone functions in L(p,q)-norms.
We continue the work of S. Tikhonov, E. Liflyand, B. Booton, and others, proving the equivalence of L(p,q)-norms of general monotone functions and of their Fourier transforms. The main tool in this work is the interpolation properties of cones of general monotone functions in L(p,q)-norms.
△ Less
Submitted 24 October, 2014;
originally announced October 2014.
