Opportunities and challenges in applying machine learning to voltammetric mechanistic studies

Bond, Alan M.; Zhang, Jie; Gundry, Luke; Kennedy, Gareth

doi:10.1016/j.coelec.2022.101009

Cited by 17 publications

(13 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Machine learning is when “computer systems are able to learn and adapt without following explicit instructions, by using algorithms and statistical models to analyse and draw inferences from patterns in data” (Bishop & Nasrabadi, 2006; Bond et al, 2022). Although machine learning has existed for several decades (the concept was first invented in the 1940s–1950s), recent advances in computing, combined with increased processing power and data availability, have enabled the approach to considerably take off (Fradkov, 2020; Goodfellow et al, 2016).…”

Section: Machine Learning and Neural Networkmentioning

confidence: 99%

Integrating machine learning, remote sensing and citizen science to create an early warning system for biodiversity

Antonelli

Dhanjal‐Adams

Silvestro

2022

Plants People Planet

View full text Add to dashboard Cite

Societal Impact Statement The development and implementation of conservation actions that address urgent threats to biodiversity across the globe require large amounts of historic and current data. Machine learning approaches offer the tools to process, analyse and interpret large data sets, giving insights into trends and guiding evidence‐based allocation of limited resources to maximise positive biodiversity outcomes. Here we describe how data acquired from remote sensing, citizen science and other monitoring approaches could feed in near‐real time into an early warning system for biodiversity that integrates automated red‐listing of species with the identification of priority areas for conservation. Summary Application of machine learning approaches is aiding biodiversity conservation and research at a time of rapid global change. Two emerging topics and their data requirements are presented. First, to identify areas of priority protection for preventing biodiversity loss, reinforcement learning is used by training models that take into account human disturbance and climate change under recurrent monitoring schemes. Second, neural networks are used to approximate classification of species into Red List categories of the International Union for Conservation of Nature, offering the possibility of real‐time re‐classification after events such as widespread fires and deforestation. We discuss how the identification of areas and species most at risk could be integrated into an ‘early warning system’ based on climatic monitoring, remotely sensed land‐use changes and near‐real time biological and threat data from citizen science initiatives. Such system would help guide actions to prevent biodiversity loss at the speed required for effective conservation.

show abstract

Section: Machine Learning and Neural Networkmentioning

confidence: 99%

Integrating machine learning, remote sensing and citizen science to create an early warning system for biodiversity

Antonelli

Dhanjal‐Adams

Silvestro

2022

Plants People Planet

View full text Add to dashboard Cite

show abstract

“…Fourth, the starting point for quantitative analysis in redox bioelectronics is electrochemical reaction–diffusion models that are commonly used for bottom-up analysis of simple electrochemical systems (e.g., to determine reaction mechanisms). − As the number of interacting redox species increases, schematics for these reaction–diffusion models begin to resemble reaction networks with the individual reacting species being network nodes and the node–node electron-transfer reactions being network links. It is envisioned that such deterministic models will become less tractable as the number of nodes (and model parameters) increases and also if the structure of the redox network is unknown (e.g., a biological redox interactome may have unknown nodes and links).…”

Section: Introductionmentioning

confidence: 99%

Spectroelectrochemical Network Measurements for Redox Bioelectronics

et al. 2023

View full text Add to dashboard Cite

Redox bioelectronics enlists electrochemical methods to connect to biology through biology’s native redox modality. The activities of this redox modality involve the exchange of electrons through redox reactions, and often, individual redox reactions are embedded within larger redox-reaction networks. Here, we examined electrodes coated with redox-active but non-conducting catechol-containing hydrogels that are emerging as important materials in redox-based bioelectronics. Previous studies have shown that electron “flow” through these catechol hydrogels involves redox reactions, and in some cases, catechol’s redox-state switching can be observed by orthogonal electrical and optical measurements. Here, we extend analysis by increasing the dimensionality of dynamic optical measurements from a single wavelength to a broader spectrum and adapt a minimal deterministic network model to reveal the intrinsic structure of this additional data. This increased dimensionality enhances our capabilities for detecting and interpreting discriminating signals, and we demonstrate these capabilities by comparing the response characteristics of conducting versus redox-active hydrogels and redox networks with different topologies. We discuss the importance of increasing measurement dimensionality to enhance both data-driven and theory-guided approaches for information processing in redox-based bioelectronics.

show abstract

“…These methods have been developed for a variety of purposes, including analyte detection (e.g., biomarkers, explosive compounds) and property quantification (e.g., estimating transport and electrochemical features). [35][36][37][38][39][40][41][42][43][44][45][46][47][48] However, these computational approaches often leverage physics-agnostic methods (e.g., support-vector machines, partial least squares regression) that are difficult to extrapolate to conditions not directly examined in the training data. [46][47][48][49] In this vein, the integration of physical models into computational voltammetry algorithms may build upon the demonstrations already present in the field to enable more powerful algorithms.…”

Section: Introductionmentioning

confidence: 99%

Leveraging graphical models to enhance in situ analyte identification via multiple voltammetric techniques

Fenton

Brushett

2023

Preprint

View full text Add to dashboard Cite

Voltammetry is a powerful analytical technique for evaluating electrochemical reactions and holds particular promise for interrogating electrolyte solutions suitable for energy storage technologies, including examining features such as state-of-charge and state-of-health. However, individual voltammetry techniques are likely to be subcomponents of broader analytical workflows that incorporate complementary methods to diagnose evolving electrolyte solutions of uncertain composition. As such, we demonstrate that jointly evaluating electrolyte solutions with distinct voltammetric modes can enhance the capabilities and sensitivities of characterization protocols. Specifically, by considering both macroelectrode cyclic square wave and microelectrode cyclic voltammograms in sequential ("one after another") and simultaneous ("all at once") manners, the composition of an electrolyte solution may be estimated with greater accuracy, and analytes that exhibit near identical electrode potentials may be more readily differentiated. We additionally explore means of further improving this method, finding that protocol accuracy increases when multiple voltammetry techniques are included in the training dataset. We also observe that the algorithm typically becomes more confident--but not necessarily more accurate--when the number of data points increases. Overall, these studies show that the sequential and simultaneous methods may hold utility when evaluating multiple voltammetry datasets that, in turn, may be leveraged to streamline diagnostic workflows used to examine electrolyte solutions within electrochemical technologies.

show abstract

Opportunities and challenges in applying machine learning to voltammetric mechanistic studies

Cited by 17 publications

References 33 publications

Integrating machine learning, remote sensing and citizen science to create an early warning system for biodiversity

Integrating machine learning, remote sensing and citizen science to create an early warning system for biodiversity

Spectroelectrochemical Network Measurements for Redox Bioelectronics

Leveraging graphical models to enhance in situ analyte identification via multiple voltammetric techniques

Contact Info

Product

Resources

About