-
The Catechol Benchmark: Time-series Solvent Selection Data for Few-shot Machine Learning
Authors:
Toby Boyne,
Juan S. Campos,
Becky D. Langdon,
Jixiang Qing,
Yilin Xie,
Shiqiang Zhang,
Calvin Tsay,
Ruth Misener,
Daniel W. Davies,
Kim E. Jelfs,
Sarah Boyall,
Thomas M. Dixon,
Linden Schrecker,
Jose Pablo Folch
Abstract:
Machine learning has promised to change the landscape of laboratory chemistry, with impressive results in molecular property prediction and reaction retro-synthesis. However, chemical datasets are often inaccessible to the machine learning community as they tend to require cleaning, thorough understanding of the chemistry, or are simply not available. In this paper, we introduce a novel dataset fo…
▽ More
Machine learning has promised to change the landscape of laboratory chemistry, with impressive results in molecular property prediction and reaction retro-synthesis. However, chemical datasets are often inaccessible to the machine learning community as they tend to require cleaning, thorough understanding of the chemistry, or are simply not available. In this paper, we introduce a novel dataset for yield prediction, providing the first-ever transient flow dataset for machine learning benchmarking, covering over 1200 process conditions. While previous datasets focus on discrete parameters, our experimental set-up allow us to sample a large number of continuous process conditions, generating new challenges for machine learning models. We focus on solvent selection, a task that is particularly difficult to model theoretically and therefore ripe for machine learning applications. We showcase benchmarking for regression algorithms, transfer-learning approaches, feature engineering, and active learning, with important applications towards solvent replacement and sustainable manufacturing.
△ Less
Submitted 9 June, 2025;
originally announced June 2025.
-
Limeade: Let integer molecular encoding aid
Authors:
Shiqiang Zhang,
Christian W. Feldmann,
Frederik Sandfort,
Miriam Mathea,
Juan S. Campos,
Ruth Misener
Abstract:
Mixed-integer programming (MIP) is a well-established framework for computer-aided molecular design (CAMD). By precisely encoding the molecular space and score functions, e.g., a graph neural network, the molecular design problem is represented and solved as an optimization problem, the solution of which corresponds to a molecule with optimal score. However, both the extremely large search space a…
▽ More
Mixed-integer programming (MIP) is a well-established framework for computer-aided molecular design (CAMD). By precisely encoding the molecular space and score functions, e.g., a graph neural network, the molecular design problem is represented and solved as an optimization problem, the solution of which corresponds to a molecule with optimal score. However, both the extremely large search space and complicated scoring process limit the use of MIP-based CAMD to specific and tiny problems. Moreover, optimal molecule may not be meaningful in practice if scores are imperfect. Instead of pursuing optimality, this paper exploits the ability of MIP in molecular generation and proposes Limeade as an end-to-end tool from real-world needs to feasible molecules. Beyond the basic constraints for structural feasibility, Limeade supports inclusion and exclusion of SMARTS patterns, automating the process of interpreting and formulating chemical requirements to mathematical constraints.
△ Less
Submitted 25 November, 2024;
originally announced November 2024.
-
Verifying message-passing neural networks via topology-based bounds tightening
Authors:
Christopher Hojny,
Shiqiang Zhang,
Juan S. Campos,
Ruth Misener
Abstract:
Since graph neural networks (GNNs) are often vulnerable to attack, we need to know when we can trust them. We develop a computationally effective approach towards providing robust certificates for message-passing neural networks (MPNNs) using a Rectified Linear Unit (ReLU) activation function. Because our work builds on mixed-integer optimization, it encodes a wide variety of subproblems, for exam…
▽ More
Since graph neural networks (GNNs) are often vulnerable to attack, we need to know when we can trust them. We develop a computationally effective approach towards providing robust certificates for message-passing neural networks (MPNNs) using a Rectified Linear Unit (ReLU) activation function. Because our work builds on mixed-integer optimization, it encodes a wide variety of subproblems, for example it admits (i) both adding and removing edges, (ii) both global and local budgets, and (iii) both topological perturbations and feature modifications. Our key technology, topology-based bounds tightening, uses graph structure to tighten bounds. We also experiment with aggressive bounds tightening to dynamically change the optimization constraints by tightening variable bounds. To demonstrate the effectiveness of these strategies, we implement an extension to the open-source branch-and-cut solver SCIP. We test on both node and graph classification problems and consider topological attacks that both add and remove edges.
△ Less
Submitted 21 May, 2024; v1 submitted 21 February, 2024;
originally announced February 2024.
-
Augmenting optimization-based molecular design with graph neural networks
Authors:
Shiqiang Zhang,
Juan S. Campos,
Christian Feldmann,
Frederik Sandfort,
Miriam Mathea,
Ruth Misener
Abstract:
Computer-aided molecular design (CAMD) studies quantitative structure-property relationships and discovers desired molecules using optimization algorithms. With the emergence of machine learning models, CAMD score functions may be replaced by various surrogates to automatically learn the structure-property relationships. Due to their outstanding performance on graph domains, graph neural networks…
▽ More
Computer-aided molecular design (CAMD) studies quantitative structure-property relationships and discovers desired molecules using optimization algorithms. With the emergence of machine learning models, CAMD score functions may be replaced by various surrogates to automatically learn the structure-property relationships. Due to their outstanding performance on graph domains, graph neural networks (GNNs) have recently appeared frequently in CAMD. But using GNNs introduces new optimization challenges. This paper formulates GNNs using mixed-integer programming and then integrates this GNN formulation into the optimization and machine learning toolkit OMLT. To characterize and formulate molecules, we inherit the well-established mixed-integer optimization formulation for CAMD and propose symmetry-breaking constraints to remove symmetric solutions caused by graph isomorphism. In two case studies, we investigate fragment-based odorant molecular design with more practical requirements to test the compatibility and performance of our approaches.
△ Less
Submitted 6 December, 2023;
originally announced December 2023.
-
Optimizing over trained GNNs via symmetry breaking
Authors:
Shiqiang Zhang,
Juan S. Campos,
Christian Feldmann,
David Walz,
Frederik Sandfort,
Miriam Mathea,
Calvin Tsay,
Ruth Misener
Abstract:
Optimization over trained machine learning models has applications including: verification, minimizing neural acquisition functions, and integrating a trained surrogate into a larger decision-making problem. This paper formulates and solves optimization problems constrained by trained graph neural networks (GNNs). To circumvent the symmetry issue caused by graph isomorphism, we propose two types o…
▽ More
Optimization over trained machine learning models has applications including: verification, minimizing neural acquisition functions, and integrating a trained surrogate into a larger decision-making problem. This paper formulates and solves optimization problems constrained by trained graph neural networks (GNNs). To circumvent the symmetry issue caused by graph isomorphism, we propose two types of symmetry-breaking constraints: one indexing a node 0 and one indexing the remaining nodes by lexicographically ordering their neighbor sets. To guarantee that adding these constraints will not remove all symmetric solutions, we construct a graph indexing algorithm and prove that the resulting graph indexing satisfies the proposed symmetry-breaking constraints. For the classical GNN architectures considered in this paper, optimizing over a GNN with a fixed graph is equivalent to optimizing over a dense neural network. Thus, we study the case where the input graph is not fixed, implying that each edge is a decision variable, and develop two mixed-integer optimization formulations. To test our symmetry-breaking strategies and optimization formulations, we consider an application in molecular design.
△ Less
Submitted 12 October, 2023; v1 submitted 16 May, 2023;
originally announced May 2023.
