A5062912204- Reinforcement Learning in Robotics
- Machine Learning and Data Classification
- Machine Learning and Algorithms
- vaccines and immunoinformatics approaches
- Gaussian Processes and Bayesian Inference
- Advanced Multi-Objective Optimization Algorithms
- Advanced Bandit Algorithms Research
- Monoclonal and Polyclonal Antibodies Research
- Polynomial and algebraic computation
- Protein purification and stability
- Robotic Path Planning Algorithms
- Advanced Numerical Analysis Techniques
- Neural Networks and Applications
- Computational Drug Discovery Methods
- Protein Structure and Dynamics
- Advanced Control Systems Optimization
- Fault Detection and Control Systems
- Autonomous Vehicle Technology and Safety
- Metaheuristic Optimization Algorithms Research
- Transportation and Mobility Innovations
- Consumer Market Behavior and Pricing
- Computational Geometry and Mesh Generation
- Language and cultural evolution
- Machine Learning in Bioinformatics
- Cancer Treatment and Pharmacology
Google (United Kingdom)
2024
DeepMind (United Kingdom)
2024
Technical University of Darmstadt
2022-2023
University College London
2018-2022
Huawei Technologies (United Kingdom)
2022
University of Cambridge
2022
Huawei Technologies (Sweden)
2022
Huawei Technologies (China)
2020
Art Institute of Portland
2019
Abstract The introduction of AlphaFold 2 1 has spurred a revolution in modelling the structure proteins and their interactions, enabling huge range applications protein design 2–6 . Here we describe our 3 model with substantially updated diffusion-based architecture that is capable predicting joint complexes including proteins, nucleic acids, small molecules, ions modified residues. new demonstrates improved accuracy over many previous specialized tools: far greater for protein–ligand...
Multi-agent interaction is a fundamental aspect of autonomous driving in the real world. Despite more than decade research and development, problem how to competently interact with diverse road users scenarios remains largely unsolved. Learning methods have much offer towards solving this problem. But they require realistic multi-agent simulator that generates competent interactions. To meet need, we develop dedicated simulation platform called SMARTS (Scalable Multi-Agent RL Training...
In this work we rigorously analyse assumptions inherent to black-box optimisation hyper-parameter tuning tasks. Our results on the Bayesmark benchmark indicate that heteroscedasticity and non-stationarity pose significant challenges for optimisers. Based these findings, propose a Heteroscedastic Evolutionary Bayesian Optimisation solver (HEBO). HEBO performs non-linear input output warping, admits exact marginal log-likelihood is robust values of learned parameters. We demonstrate HEBO’s...
Antibodies are multimeric proteins capable of highly specific molecular recognition. The complementarity determining region 3 the antibody variable heavy chain (CDRH3) often dominates antigen-binding specificity. Hence, it is a priority to design optimal antigen-specific CDRH3 develop therapeutic antibodies. combinatorial structure sequences makes impossible query binding-affinity oracles exhaustively. Moreover, antibodies expected have high target specificity and developability. Here, we...
We introduce a method combining variational autoencoders (VAEs) and deep metric learning to perform Bayesian optimisation (BO) over high-dimensional structured input spaces. By adapting ideas from learning, we use label guidance the blackbox function structure VAE latent space, facilitating Gaussian process fit yielding improved BO performance. Importantly for problem settings, our operates in semi-supervised regimes where only few labelled data points are available. run experiments on three...
In this work we rigorously analyse assumptions inherent to black-box optimisation hyper-parameter tuning tasks. Our results on the Bayesmark benchmark indicate that heteroscedasticity and non-stationarity pose significant challenges for optimisers. Based these findings, propose a Heteroscedastic Evolutionary Bayesian Optimisation solver (HEBO). HEBO performs non-linear input output warping, admits exact marginal log-likelihood is robust values of learned parameters. We demonstrate HEBO's...
Bayesian optimisation presents a sample-efficient methodology for global optimisation. Within this framework, crucial performance-determining subroutine is the maximisation of acquisition function, task complicated by fact that functions tend to be non-convex and thus nontrivial optimise. In paper, we undertake comprehensive empirical study approaches maximise function. Additionally, deriving novel, yet mathematically equivalent, compositional forms popular functions, recast as problem,...
Antibodies are canonically Y-shaped multimeric proteins capable of highly specific molecular recognition. The CDRH3 region located at the tip variable chains an antibody dominates antigen-binding specificity. Therefore, it is a priority to design optimal antigen-specific regions develop therapeutic antibodies. However, combinatorial nature sequence space makes impossible search for binding exhaustively and efficiently using computational approaches. Here, we present \texttt{AntBO}:...
Satisfying safety constraints almost surely (or with probability one) can be critical for the deployment of Reinforcement Learning (RL) in real-life applications. For example, plane landing and take-off should ideally occur one. We address problem by introducing Safety Augmented (Saute) Markov Decision Processes (MDPs), where are eliminated augmenting them into state-space reshaping objective. show that Saute MDP satisfies Bellman equation moves us closer to solving Safe RL satisfied surely....
Recent advances in Neural Variational Inference allowed for a renaissance latent variable models variety of domains involving high-dimensional data. While traditional variational methods derive an analytical approximation the intractable distribution over variables, here we construct inference network conditioned on symbolic representation entities and relation types Knowledge Graph, to provide distributions. The new framework results highly-scalable method. Under Bernoulli sampling...
Optimizing combinatorial structures is core to many real-world problems, such as those encountered in life sciences. For example, one of the crucial steps involved antibody design find an arrangement amino acids a protein sequence that improves its binding with pathogen. Combinatorial optimization antibodies difficult due extremely large search spaces and non-linear objectives. Even for modest where proteins have length eleven, we are faced searching over 2.05 x 10^14 structures. Applying...
Cylindrical Algebraic Decomposition (CAD) is an important tool within computational real algebraic geometry, capable of solving many problems to do with polynomial systems over the reals, but known have worst-case complexity doubly exponential in number variables. It has long been studied by Symbolic Computation community and implemented a variety computer algebra systems, however, it also found recent interest Satisfiability Checking for use SMT-solvers. The SCSC Project seeks build bridges...
Cylindrical Algebraic Decomposition (CAD) is an important tool within computational real algebraic geometry, capable of solving many problems for polynomial systems over the reals. It has long been studied by Symbolic Computation community and found recent interest in Satisfiability Checking community. The present report describes a proof concept implementation Incremental CAD algorithm Maple, where CADs are built then refined as additional constraints added. aim to make suitable use theory...
Real-time bidding (RTB) is a popular method to sell online ad space inventory using real-time auctions determine which advertiser gets make the impression. Advertisers can take user information into account when making their bids and get more control over process. The goal of an optimal function maximise overall effectiveness campaigns defined by advertisers under certain budget constraint. A straightforward solution would be model in explicit form. However, such functional solutions lack...
Searching for bindings of geometric parameters in task and motion planning (TAMP) is a finite-horizon stochastic problem with high-dimensional decision spaces. A robot manipulator can only move subspace its whole range that subjected to many constraints. TAMP solver usually takes explorations before finding feasible binding set each task. It favorable learn those constraints once then transfer them over different tasks within the same workspace. We address this by representing constraint...
Antibodies are canonically Y-shaped multimeric proteins capable of highly specific molecular recognition. The CDRH3 region located at the tip variable chains an antibody dominates antigen-binding specificity. Therefore, it is a priority to design optimal antigen-specific regions develop therapeutic antibodies. However, combinatorial nature sequence space makes impossible search for binding exhaustively and efficiently using computational approaches. Here, we present \texttt{AntBO}: Bayesian...