The predictability of a lake phytoplankton community, from hours to years

AbstractForecasting anthropogenic changes to ecological communities is one of the central challenges in ecology. However, nonlinear dependencies, biotic interactions and data limitations have limited our ability to assess how predictable communities are. Here we used a machine learning approach and environmental monitoring data (biological, physical and chemical) to assess the predictability of phytoplankton cell density in one lake across an unprecedented range of time scales. Communities were highly predictable over hours to months: model R2 decreased from 0. 89 at 4 hours to 0.75 at 1 month, and in a long-term dataset lacking fine spatial resolution, from 0.46 at 1 month to 0.32 at 10 years. When cyanobacterial and eukaryotic algal cell density were examined separately, model-inferred environmental growth dependencies matched laboratory studies, and suggested novel trade-offs governing their competition. High-frequency monitoring and machine learning can help elucidate the mechanisms underlying ecological dynamics and set prediction targets for process-based models.

Download Full-text

Predicting Undesired Treatment Outcome in Mental Healthcare: Machine Learning Study (Preprint)

10.2196/preprints.17235 ◽

2019 ◽

Author(s):

Kasper Van Mens ◽

Joran Lokkerbol ◽

Richard Janssen ◽

Robert de Lange ◽

Bea Tiemens

Keyword(s):

Machine Learning ◽

Treatment Outcome ◽

Mental Health Treatment ◽

Mental Healthcare ◽

Machine Learning Algorithms ◽

Gradient Boosting ◽

Trade Off ◽

Trade Offs ◽

Outcome Monitoring ◽

Extreme Gradient Boosting

BACKGROUND It remains a challenge to predict which treatment will work for which patient in mental healthcare. OBJECTIVE In this study we compare machine algorithms to predict during treatment which patients will not benefit from brief mental health treatment and present trade-offs that must be considered before an algorithm can be used in clinical practice. METHODS Using an anonymized dataset containing routine outcome monitoring data from a mental healthcare organization in the Netherlands (n = 2,655), we applied three machine learning algorithms to predict treatment outcome. The algorithms were internally validated with cross-validation on a training sample (n = 1,860) and externally validated on an unseen test sample (n = 795). RESULTS The performance of the three algorithms did not significantly differ on the test set. With a default classification cut-off at 0.5 predicted probability, the extreme gradient boosting algorithm showed the highest positive predictive value (ppv) of 0.71(0.61 – 0.77) with a sensitivity of 0.35 (0.29 – 0.41) and area under the curve of 0.78. A trade-off can be made between ppv and sensitivity by choosing different cut-off probabilities. With a cut-off at 0.63, the ppv increased to 0.87 and the sensitivity dropped to 0.17. With a cut-off of at 0.38, the ppv decreased to 0.61 and the sensitivity increased to 0.57. CONCLUSIONS Machine learning can be used to predict treatment outcomes based on routine monitoring data.This allows practitioners to choose their own trade-off between being selective and more certain versus inclusive and less certain.

Download Full-text

Abnormal identification of lubricating oil parameters and evaluation of physical and chemical properties based on machine learning

IOP Conference Series Materials Science and Engineering ◽

10.1088/1757-899x/1043/5/052053 ◽

2021 ◽

Vol 1043 (5) ◽

pp. 052053

Author(s):

Kun Yang ◽

Xia Wang

Keyword(s):

Machine Learning ◽

Chemical Properties ◽

Lubricating Oil ◽

Physical And Chemical Properties ◽

Physical And Chemical ◽

And Chemical Properties

Download Full-text

Ethical Adversaries

ACM SIGKDD Explorations Newsletter ◽

10.1145/3468507.3468513 ◽

2021 ◽

Vol 23 (1) ◽

pp. 32-41

Author(s):

Pieter Delobelle ◽

Paul Temple ◽

Gilles Perrouin ◽

Benoit Frénay ◽

Patrick Heymans ◽

...

Keyword(s):

Machine Learning ◽

Equality Of Opportunity ◽

Decision Processes ◽

Critical Systems ◽

Trade Offs ◽

Theoretical Considerations

Machine learning is being integrated into a growing number of critical systems with far-reaching impacts on society. Unexpected behaviour and unfair decision processes are coming under increasing scrutiny due to this widespread use and its theoretical considerations. Individuals, as well as organisations, notice, test, and criticize unfair results to hold model designers and deployers accountable. We offer a framework that assists these groups in mitigating unfair representations stemming from the training datasets. Our framework relies on two inter-operating adversaries to improve fairness. First, a model is trained with the goal of preventing the guessing of protected attributes' values while limiting utility losses. This first step optimizes the model's parameters for fairness. Second, the framework leverages evasion attacks from adversarial machine learning to generate new examples that will be misclassified. These new examples are then used to retrain and improve the model in the first step. These two steps are iteratively applied until a significant improvement in fairness is obtained. We evaluated our framework on well-studied datasets in the fairness literature - including COMPAS - where it can surpass other approaches concerning demographic parity, equality of opportunity and also the model's utility. We investigated the trade-offs between these targets in terms of model hyperparameters and also illustrated our findings on the subtle difficulties when mitigating unfairness and highlight how our framework can assist model designers.

Download Full-text

Lessons learned from data-limited evaluations of data-rich reef fish species in the Gulf of Mexico: implications for providing fisheries management advice for data-poor stocks

Canadian Journal of Fisheries and Aquatic Sciences ◽

10.1139/cjfas-2017-0482 ◽

2019 ◽

Vol 76 (9) ◽

pp. 1624-1639 ◽

Cited By ~ 3

Author(s):

Skyler R. Sagarese ◽

William J. Harford ◽

John F. Walter ◽

Meaghan D. Bryan ◽

J. Jeffery Isely ◽

...

Keyword(s):

Gulf Of Mexico ◽

Reef Fish ◽

Fish Species ◽

Stock Assessment ◽

Lessons Learned ◽

Management Objectives ◽

Trade Offs ◽

Data Limitations ◽

Management Procedures ◽

Reef Fish Species

Specifying annual catch limits for artisanal fisheries, low economic value stocks, or bycatch species is problematic due to data limitations. Many empirical management procedures (MPs) have been developed that provide catch advice based on achieving a stable catch or a historical target (i.e., instead of maximum sustainable yield). However, a thorough comparison of derived yield streams between empirical MPs and stock assessment models has not been explored. We first evaluate trade-offs in conservation and yield metrics for data-limited approaches through management strategy evaluation (MSE) of seven data-rich reef fish species in the Gulf of Mexico. We then apply data-limited approaches for each species and compare how catch advice differs from current age-based assessment models. MSEs identified empirical MPs (e.g., using relative abundance) as a compromise between data requirements and the ability to consistently achieve management objectives (e.g., prevent overfishing). Catch advice differed greatly among data-limited approaches and current assessments, likely due to data inputs and assumptions. Adaptive MPs become clearly viable options that can achieve management objectives while incorporating auxiliary data beyond catch-only approaches.

Download Full-text

Work Flows for Cellular Epidemiology, From Conception to Translation

10.1101/548412 ◽

2019 ◽

Author(s):

Anton D. Nathanson ◽

Lucy Ngo ◽

Tomasz Garbowski ◽

Abhilash Srikantha ◽

Christian Wojek ◽

...

Keyword(s):

Electron Microscopy ◽

Machine Learning ◽

Image Resolution ◽

Machine Learning Algorithms ◽

Human Anatomy ◽

Test Case ◽

Health State ◽

Learning Approaches ◽

Preparation Methods ◽

Trade Offs

AbstractChanges in cell connectivity and morphology, observed and measured using microscopy, implicate a cellular basis of degenerative disease in tissues as diverse as bone, kidney and brain. To date, limitations inherent to sampling (biopsy sites) and/or microscopy (trade-offs between regions of interest and image resolution) have prevented early identification of cellular changes in specimen sizes of diagnostic relevance for human anatomy and physiology. This manuscript describes work flows for human tissue-based cell epidemiology studies. Using recently published sample preparation methods, developed and validated to maximize imaging quality, the largest-to-date scanning electron microscopy map was created showing cellular connections in the femoral neck of a human hip. The map, from a patient undergoing hip replacement, comprises an 11 TB dataset including over 7 million electron microscopy images. This map served as a test case to implement machine learning algorithms for automated detection of cells and identification of their health state. The test case showed a significant link between cell connectivity and health state in osteocytes of the human femur. Combining new, rapid throughput electron microscopy methods with machine learning approaches provides a basis for assessment of cell population health at nanoscopic resolution and in mesoscopic tissue and organ samples. This sets a path for next generation cellular epidemiology, tracking outbreaks of disease in populations of cells that inhabit tissues and organs within individuals.

Download Full-text

Open-Ended Questions

Employee Surveys and Sensing ◽

10.1093/oso/9780190939717.003.0013 ◽

2020 ◽

pp. 202-218

Author(s):

Subhadra Dutta ◽

Eric M. O’Rourke

Keyword(s):

Machine Learning ◽

Language Processing ◽

Latent Dirichlet Allocation ◽

Semantic Analysis ◽

Written Language ◽

Text Data ◽

Employee Survey ◽

Trade Offs ◽

Word Relatedness ◽

Survey Responses

Natural language processing (NLP) is the field of decoding human written language. This chapter responds to the growing interest in using machine learning–based NLP approaches for analyzing open-ended employee survey responses. These techniques address scalability and the ability to provide real-time insights to make qualitative data collection equally or more desirable in organizations. The chapter walks through the evolution of text analytics in industrial–organizational psychology and discusses relevant supervised and unsupervised machine learning NLP methods for survey text data, such as latent Dirichlet allocation, latent semantic analysis, sentiment analysis, word relatedness methods, and so on. The chapter also lays out preprocessing techniques and the trade-offs of growing NLP capabilities internally versus externally, points the readers to available resources, and ends with discussing implications and future directions of these approaches.

Download Full-text

The role of biotic interactions in invasion ecology: theories and hypotheses.

Plant invasions: the role of biotic interactions ◽

10.1079/9781789242171.0002 ◽

2020 ◽

pp. 26-44

Author(s):

Cang Hui ◽

◽

Pietro Landi ◽

Guillaume Latombe ◽

◽

...

Keyword(s):

Native Species ◽

Evolutionary Dynamics ◽

Biotic Interactions ◽

Evolutionary Games ◽

Invasion Ecology ◽

Ecological Communities ◽

Dynamic Interactions ◽

Ecological Fitting ◽

And Function

Changes in biotic interactions in the native and invaded range can enable a non-native species to establish and spread in novel environments. Invasive non-native species can in turn generate impacts in recipient systems partly through the changes they impose on biotic interactions; these interactions can lead to altered ecosystem processes in the recipient systems. This chapter reviews models, theories and hypotheses on how invasion performance and impact of introduced species in recipient ecosystems can be conjectured according to biotic interactions between native and non-native species. It starts by exploring the nature of biotic interactions as ensembles of ecological and evolutionary games between individuals of both the same and different groups. This allows us to categorize biotic interactions as direct and indirect (i.e. those involving more than two species) that emerge from both coevolution and ecological fitting during community assembly and invasion. We then introduce conceptual models that can reveal the ecological and evolutionary dynamics between interacting non-native and resident species in ecological networks and communities. Moving from such theoretical grounding, we review 20 hypotheses that have been proposed in invasion ecology to explain the invasion performance of a single non-native species, and seven hypotheses relating to the creation and function of assemblages of non-native species within recipient ecosystems. We argue that, although biotic interactions are ubiquitous and quintessential to the assessment of invasion performance, they are nonetheless difficult to detect and measure due to strength dependency on sampling scales and population densities, as well as the non-equilibrium transient dynamics of ecological communities and networks. We therefore call for coordinated efforts in invasion science and beyond, to devise and review approaches that can rapidly map out the entire web of dynamic interactions in a recipient ecosystem.

Download Full-text

Quasi-Operational Testing of Real-Time Storm-Longevity Prediction via Machine Learning

Weather and Forecasting ◽

10.1175/waf-d-18-0141.1 ◽

2019 ◽

Vol 34 (5) ◽

pp. 1437-1451 ◽

Cited By ~ 2

Author(s):

Amy McGovern ◽

Christopher D. Karstens ◽

Travis Smith ◽

Ryan Lagerquist

Keyword(s):

Machine Learning ◽

Real Time ◽

Weather Prediction ◽

Warning System ◽

Objective Evaluation ◽

Boosted Regression Trees ◽

Time Prediction ◽

Operational Testing ◽

Trade Offs ◽

Naturalistic Environment

Abstract Real-time prediction of storm longevity is a critical challenge for National Weather Service (NWS) forecasters. These predictions can guide forecasters when they issue warnings and implicitly inform them about the potential severity of a storm. This paper presents a machine-learning (ML) system that was used for real-time prediction of storm longevity in the Probabilistic Hazard Information (PHI) tool, making it a Research-to-Operations (R2O) project. Currently, PHI provides forecasters with real-time storm variables and severity predictions from the ProbSevere system, but these predictions do not include storm longevity. We specifically designed our system to be tested in PHI during the 2016 and 2017 Hazardous Weather Testbed (HWT) experiments, which are a quasi-operational naturalistic environment. We considered three ML methods that have proven in prior work to be strong predictors for many weather prediction tasks: elastic nets, random forests, and gradient-boosted regression trees. We present experiments comparing the three ML methods with different types of input data, discuss trade-offs between forecast quality and requirements for real-time deployment, and present both subjective (human-based) and objective evaluation of real-time deployment in the HWT. Results demonstrate that the ML system has lower error than human forecasters, which suggests that it could be used to guide future storm-based warnings, enabling forecasters to focus on other aspects of the warning system.

Download Full-text