Parameter-rich Models Research Articles

A simple but tough-to-beat baseline for fMRI time-series classification

Current neuroimaging studies frequently use complex machine learning models to classify human fMRI data, distinguishing healthy and disordered brains, often to validate new methods or enhance prediction accuracy. Yet, where prediction accuracy is a concern, our results suggest that precision in prediction does not always require such sophistication. When a classifier as simple as logistic regression is applied to feature-engineered fMRI data, it can match or even outperform more sophisticated recent models. Classification of the raw time series fMRI data generally benefits from complex parameter-rich models. However, this complexity often pushes them into the class of black-box models. Yet, we found that a relatively simple model can consistently outperform much more complex classifiers in both accuracy and speed. This model applies the same multi-layer perceptron repeatedly across time and averages the results. Thus, the complexity and black-box nature of the parameter rich models, often perceived as a necessary trade-off for higher performance, do not invariably yield superior results on fMRI.Given the success of straightforward approaches, we challenge the merit of research that concentrates solely on complex model development driven by classification. Instead, we advocate for increased focus on designing models that prioritize the explainability of fMRI data or pursue applicable objectives beyond mere classification accuracy, unless they significantly outperform logistic regression or our proposed model. To validate our claim, we explore possible reasons for the superior performance of our straightforward model by examining the innate characteristics of fMRI time series data. Our findings suggest that the sequential information hidden in the temporal order may be far less important for the accurate fMRI classification than the stand-alone pieces of information scattered across the frames of the time series.

Spectral cluster supertree: fast and statistically robust merging of rooted phylogenetic trees.

The algorithms for phylogenetic reconstruction are central to computational molecular evolution. The relentless pace of data acquisition has exposed their poor scalability and the conclusion that the conventional application of these methods is impractical and not justifiable from an energy usage perspective. Furthermore, the drive to improve the statistical performance of phylogenetic methods produces increasingly parameter-rich models of sequence evolution, which worsens the computational performance. Established theoretical and algorithmic results identify supertree methods as critical to divide-and-conquer strategies for improving scalability of phylogenetic reconstruction. Of particular importance is the ability to explicitly accommodate rooted topologies. These can arise from the more biologically plausible non-stationary models of sequence evolution. We make a contribution to addressing this challenge with Spectral Cluster Supertree, a novel supertree method for merging a set of overlapping rooted phylogenetic trees. It offers significant improvements over Min-Cut supertree and previous state-of-the-art methods in terms of both time complexity and overall topological accuracy, particularly for problems of large size. We perform comparisons against Min-Cut supertree and Bad Clade Deletion. Leveraging two tree topology distance metrics, we demonstrate that while Bad Clade Deletion generates more correct clades in its resulting supertree, Spectral Cluster Supertree's generated tree is generally more topologically close to the true model tree. Over large datasets containing 10,000 taxa and 500 source trees, where Bad Clade Deletion usually takes 2 h to run, our method generates a supertree in on average 20 s. Spectral Cluster Supertree is released under an open source license and is available on the python package index as sc-supertree.

Stochastic heuristics for decisions under risk and uncertainty.

Models of heuristics are often predicated on the desideratum that they should possess no free parameters. As a result, heuristic implementations are usually deterministic and do not allow for any choice errors, as the latter would require a parameter to regulate the magnitude of errors. We discuss the implications of this in light of research that highlights the evidence supporting stochastic choice and its dependence on preferential strength. We argue that, in principle, the existing models of deterministic heuristics should, and can, be quite easily modified to stochastic counterparts through the addition of an error mechanism. This requires a single free parameter in the error mechanism, whilst otherwise retaining the parameter-free cognitive processes in the deterministic component of existing heuristics. We present various types of error mechanisms applicable to heuristics and discuss their comparative virtues and drawbacks, paying particular attention to their impact on model comparisons between heuristics and parameter-rich models.

Organellar phylogenomics at the epidendroid orchid base, with a focus on the mycoheterotrophic Wullschlaegelia.

Heterotrophic plants have long been a challenge for systematists, exemplified by the base of the orchid subfamily Epidendroideae, which contains numerous mycoheterotrophic species. Here we address the utility of organellar genomes in resolving relationships at the epidendroid base, specifically employing models of heterotachy, or lineage-specific rate variation over time. We further conduct comparative analyses of plastid genome evolution in heterotrophs and structural variation in matK. We present the first complete plastid genomes (plastomes) of Wullschlaegelia, the sole genus of the tribe Wullschlaegelieae, revealing a highly reduced genome of 37 kilobases, which retains a fraction of the genes present in related autotrophs. Plastid phylogenomic analyses recovered a strongly supported clade composed exclusively of mycoheterotrophic species with long branches. We further analyzed mitochondrial gene sets, which recovered similar relationships to those in other studies using nuclear data, but the placement of Wullschlaegelia remains uncertain. We conducted comparative plastome analyses among Wullschlaegelia and other heterotrophic orchids, revealing a suite of correlated substitutional and structural changes relative to autotrophic species. Lastly, we investigated evolutionary and structural variation in matK, which is retained in Wullschlaegelia and a few other 'late stage' heterotrophs and found evidence for structural conservation despite rapid substitution rates in both Wullschlaegelia and the leafless Gastrodia. Our analyses reveal the limits of what the plastid genome can tell us on orchid relationships in this part of the tree, even when applying parameter-rich heterotachy models. Our study underscores the need for increased taxon sampling across all three genomes at the epidendroid base, and illustrates the need for further research on addressing heterotachy in phylogenomic analyses.

Unstable cores are the source of instability in chemical reaction networks

In biochemical networks, complex dynamical features such as superlinear growth and oscillations are classically considered a consequence of autocatalysis. For the large class of parameter-rich kinetic models, which includes generalized mass action kinetics and Michaelis–Menten kinetics, we show that certain submatrices of the stoichiometric matrix, so-called unstable cores, are sufficient for a reaction network to admit instability and potentially give rise to such complex dynamical behaviour. The determinant of the submatrix distinguishes unstable-positive feedbacks, with a single real-positive eigenvalue, and unstable-negative feedbacks without real-positive eigenvalues. Autocatalytic cores turn out to be exactly the unstable-positive feedbacks that are Metzler matrices. Thus there are sources of dynamical instability in chemical networks that are unrelated to autocatalysis. We use such intuition to design non-autocatalytic biochemical networks with superlinear growth and oscillations.

Signatures of natural selection in tree topology shape of serially sampled viral phylogenies

Tree shape metrics can be computed fast for trees of any size, which makes them promising alternatives to intensive statistical methods and parameter-rich evolutionary models in the era of massive data availability. Previous studies have demonstrated their effectiveness in unveiling important parameters in viral evolutionary dynamics, although the impact of natural selection on the shape of tree topologies has not been thoroughly investigated. We carried out a forward-time and individual-based simulation to investigate whether tree shape metrics of several kinds could predict the selection regime employed to generate the data. To examine the impact of the genetic diversity of the founder viral population, simulations were run under two opposing starting configurations of the genetic diversity of the infecting viral population. We found that four evolutionary regimes, namely, negative, positive, and frequency-dependent selection, as well as neutral evolution, were successfully distinguished by tree topology shape metrics. Two metrics from the Laplacian spectral density profile (principal eigenvalue and peakedness) and the number of cherries were the most informative for indicating selection type. The genetic diversity of the founder population had an impact on differentiating evolutionary scenarios. Tree imbalance, which has been frequently associated with the action of natural selection on intrahost viral diversity, was also characteristic of neutrally evolving serially sampled data. Metrics calculated from empirical analysis of HIV datasets indicated that most tree topologies exhibited shapes closer to the frequency-dependent selection or neutral evolution regimes.

Comparison of Different Methods in Estimating Reference Evapotranspiration at Gaya District of Bihar, India

Evapotranspiration (ET), a complex mechanism in the hydrological cycle, influences runoff, which in turn influences how much water is required for irrigation. There are many techniques available to calculate of the reference evapotranspiration (ETo). Unfortunately, most efficient reference evapotranspiration approaches are parameter-rich models, making them unsuitable for use in data-scarce locations. On the other hand, accuracy and reliability of simple ETo models vary widely according to regional climate conditions. The purpose of this study was to assess the effectiveness of two temperature-based and three radiation-based simple ET methods in estimating ET algorithms used to forecast the wheat crop for the years 2017–2018 in Gaya District, Bihar. The performance was measured by comparing the temperature and radiation-based method with the parameter intensive Penman-Monteith method. Comparative evaluation of two method (radiation and temperature) was performed through statistical tests, and it was found that the (R2 = 1 and R2 = 0.98) and RMSE = 0.84. The study found that ETo values calculated from the six methods were highly correlated (Pearson correlation coefficient 0.84 to 0.99). Through data analysis for model comparison, it was found that the Radiation method performed satisfactorily when compared to the standard Penman-Monteith model estimate (R2 = 0.94). However, multivariate statistical testing revealed that the ETo readings obtained using various techniques differed significantly from one another. In comparison to radiation-based ETo approaches, temperature-based ETo techniques showed greater disparities. The Priestley-Taylor, Turc, and Jensen-Haise methods all outperformed the other ETo methods in general. When estimating ETo in this study region, it was discovered that radiation-based approaches performed better than temperature-based methods.

Importance of the Long-Term Seasonal Component in Day-Ahead Electricity Price Forecasting Revisited: Parameter-Rich Models Estimated via the LASSO

Recent studies suggest that decomposing a series of electricity spot prices into a trend-seasonal and a stochastic component, modeling them independently, and then combining their forecasts can yield more accurate predictions than an approach in which the same parsimonious regression or neural network-based model is calibrated to the prices themselves. Here, we show that significant accuracy gains can also be achieved in the case of parameter-rich models estimated via the least absolute shrinkage and selection operator (LASSO). Moreover, we provide insights as to the order of applying seasonal decomposition and variance stabilizing transformations before model calibration, and propose two well-performing forecast averaging schemes that are based on different approaches for modeling the long-term seasonal component.

Assessing the accuracy and robustness of a process-based model for coffee agroforestry systems in Central America

Coffee is often grown in production systems associated with shade trees that provide different ecosystem services. Management, weather and soil conditions are spatially variable production factors. CAF2007 is a dynamic model for coffee agroforestry systems that takes these factors as inputs and simulates the processes underlying berry production at the field scale. There remain, however, uncertainties about process rates that need to be reduced through calibration. Bayesian statistics using Markov chain Monte Carlo algorithms is increasingly used for calibration of parameter-rich models. However, very few studies have employed multi-site calibration, which aims to reduce parameter uncertainties using data from multiple sites simultaneously. The main objectives of this study were to calibrate the coffee agroforestry model using data gathered in long-term experiments in Costa Rica and Nicaragua, and to test the calibrated model against independent data from commercial coffee-growing farms. Two sub-models were improved: calculation of flowering date and the modelling of biennial production patterns. The modified model, referred to as CAF2014, can be downloaded at https://doi.org/10.5281/zenodo.3608877 . Calibration improved model performance (higher R2, lower RMSE) for Turrialba (Costa Rica) and Masatepe (Nicaragua), including when all experiments were pooled together. Multi-site and single-site Bayesian calibration led to similar RMSE. Validation on new data from coffee-growing farms revealed that both calibration methods improved simulation of yield and its bienniality. The thus improved model was used to test the effect of N fertilizer and shade in different locations on coffee yield.

A Novel Approach to Investigate the Effect of Tree Reconstruction Artifacts in Single-Gene Analysis Clarifies Opsin Evolution in Nonbilaterian Metazoans.

Our ability to correctly reconstruct a phylogenetic tree is strongly affected by both systematic errors and the amount of phylogenetic signal in the data. Current approaches to tackle tree reconstruction artifacts, such as the use of parameter-rich models, do not translate readily to single-gene alignments. This, coupled with the limited amount of phylogenetic information contained in single-gene alignments, makes gene trees particularly difficult to reconstruct. Opsin phylogeny illustrates this problem clearly. Opsins are G-protein coupled receptors utilized in photoreceptive processes across Metazoa and their protein sequences are roughly 300 amino acids long. A number of incongruent opsin phylogenies have been published and opsin evolution remains poorly understood. Here, we present a novel approach, the canary sequence approach, to investigate and potentially circumvent errors in single-gene phylogenies. First, we demonstrate our approach using two well-understood cases of long-branch attraction in single-gene data sets, and simulations. After that, we apply our approach to a large collection of well-characterized opsins to clarify the relationships of the three main opsin subfamilies.

Phylogenetic Trees and Networks Can Serve as Powerful and Complementary Approaches for Analysis of Genomic Data.

Genomic data have had a profound impact on nearly every biological discipline. In systematics and phylogenetics, the thousands of loci that are now being sequenced can be analyzed under the multispecies coalescent model (MSC) to explicitly account for gene tree discordance due to incomplete lineage sorting (ILS). However, the MSC assumes no gene flow post divergence, calling for additional methods that can accommodate this limitation. Explicit phylogenetic network methods have emerged, which can simultaneously account for ILS and gene flow by representing evolutionary history as a directed acyclic graph. In this point of view, we highlight some of the strengths and limitations of phylogenetic networks and argue that tree-based inference should not be blindly abandoned in favor of networks simply because they represent more parameter rich models. Attention should be given to model selection of reticulation complexity, and the most robust conclusions regarding evolutionary history are likely obtained when combining tree- and network-based inference.

Phylogeny, ancestors, and anagenesis in the hominin fossil record

AbstractProbabilistic approaches to phylogenetic inference have recently gained traction in paleontological studies. Because they directly model processes of evolutionary change, probabilistic methods facilitate a deeper assessment of variability in evolutionary patterns by weighing evidence for competing models. Although phylogenetic methods used in paleontological studies have generally assumed that evolution proceeds by splitting cladogenesis, extensions to previous models help explore the potential for morphological and temporal data to provide differential support for contrasting modes of evolutionary divergence. Recent methodological developments have integrated ancestral relationships into probabilistic phylogenetic methods. These new approaches rely on parameter-rich models and sophisticated inferential methods, potentially obscuring the respective contributions of data and models. In this study, we describe a simple likelihoodist approach that combines probabilistic models of morphological evolution and fossil preservation to reconstruct both cladogenetic and anagenetic relationships. By applying this approach to a data set of fossil hominins, we demonstrate the capability of existing models to unveil evidence for anagenesis presented by morphological and temporal data. This evidence was previously recognized by qualitative assessments, but largely ignored by quantitative phylogenetic analyses. For example, we find support for directly ancestral relationships in multiple lineages:Sahelanthropusis ancestral to later hominins;Australopithecus anamensisis ancestral toAustralopithecus afarensis;Australopithecus garhiis ancestral toHomo;Homo antecessoris ancestral toHomo heidelbergensis, which in turn is ancestral to bothHomo sapiensandHomo neanderthalensis. By accommodating direct ancestry in phylogenetics, quantitative results align more closely with previous qualitative expectations.

Modelling reference evapotranspiration for Megecha catchment by multiple linear regression

Reference evapotranspiration (ETo) is a complex process in the hydrological cycle that influences the quantity of runoff and then the irrigation water requirements. Many methods have been developed to estimate reference evapotranspiration. The reliable reference evapotranspiration methods are parameter rich models but not applicable for data scarce regions. On the other hand, accuracy and reliability of simple reference evapotranspiration models vary widely according to regional climate conditions. Estimation of reference evapotranspiration of specific area is imperative for investigating, planning, designing, and managing irrigation schemes. The objective of the present study is to determine the ETo values using combination Penman–Monteith model and to devise a model used to predict reference evapotranspiration in the Megecha catchment. In this study, the potential of multiple linear regression model using least squares is investigated in modeling of mean monthly reference evapotranspiration obtained using the standard FAO-56 Penman–Monteith equation. Various combinations of daily climatic data, namely solar radiation, air temperature, relative humidity, and wind speed were used as inputs to the model development. Thus, this can give direction to evaluate the degree and effect of each variable on estimated ETo. In line to this, residual analysis, co-efficient of determination and residual error were used as comparison tools for the evaluation of the model performance. The result of the multiple correlation showed that sun hour, wind speed, and maximum temperature were strongly positive correlation with ETOr = 0.82, 0.71, and 0.78, respectively, whereas relative humidity has moderately negative correlation (r = − 0.604). The multiple linear regression model gave residual errors of 0.26 mm/day and co-efficient of determination of 0.92 for the meteorological station considered. Based on the residual analysis results, it was found that the multiple linear regression model followed a linear trend. The input variables of the model were fitting, and this model could be successfully employed in estimating and predicting the monthly reference evapotranspiration in the study area and can be used in similar parts of data sparse regions.

Emulation of environmental models using polynomial chaos expansion

This paper investigates the applicability of model emulation to speed up simulation time of CPU intensive environmental models. Polynomial chaos expansion (PCE) emulators are constructed for three case studies of increasing complexity. The level of emulator training and the order of polynomial necessary to sufficiently build accurate emulators for each model are investigated. Although the PCE emulators shown here do not approximate well the outputs of parameter rich models (80 + parameters), results demonstrate that the emulators mimic closely outputs of relatively simple, low dimensional, simulation models (15 parameters or less). Furthermore, the PCE emulators are tested with applications such as Global Sensitivity Analysis (GSA). Results illustrate the advantages and drawbacks of using classical PCE emulators for treating computational limitation of complex environmental models.

Modelling German Electricity Wholesale Spot Prices with a Parsimonious Fundamental Model Validation & Application

Increasing shares of fluctuating renewable energy, the integration of European electricity grids and markets as well as new technologies induce continuous change in the European energy system. Due to these changes, fundamental electricity system and market models that have been developed and applied in the past are dealing with an increasing number of details inducing correspondingly huge data needs. The complexity of these called parameter-rich models (cf. Weron, 2014) leads to limited transparency, also on the impact of data on results, and makes model backtesting rather cumbersome. At the same time, the validity of future scenarios based on non-validated models is dubious. To complement these highly complicated models, more reduced models may be helpful both for transparency and for backtesting. In this paper, we apply a parsimonious fundamental modelling approach to determine hourly German day-ahead power market prices and production volumes. The methodology approximates the supply stack by a piecewise linear function and considers fundamental information, e.g. power plant capacities and availabilities, fuel prices, must-run production and cross-border exchange. We reduce complexity by considering technology classes, uncoupled time periods and only one market area. Between 2011 and 2015, German day-ahead prices declined by 38% and various reasons have been identified in literature, namely a drop in emission certificate prices, the expansion of renewable energies (RES) or lower fuel prices. However, the decision of the German government to shut down nuclear power plants after the Fukushima nuclear disaster happened at the same time and received too little attention as it rather by itself could have led to an increase in prices. The parsimonious model is able to reproduce the hourly historical prices (2011-2015) with a MAE of 5.6 €/MWh and accurately reproduces the electricity production volumes for most thermal production units. In a case study, we investigate a counterfactual scenario without accelerated nuclear phase-out in Germany after the Fukushima nuclear disaster in 2011. The results indicate that German day-ahead power prices would have fallen by additional 3 €/MWh if the nuclear phase-out would have not occurred. Since coal- and gas-fired production as well as additional imports have substituted production from nuclear power plants, their usage would have dropped in the counterfactual scenario.

English

A number of methods have been developed to estimate reference crop evapotranspiration (ETo), but most of the reliable methods are complex and parameter rich models and therefore difficult to apply in data scarce regions. This study was conducted to determine suitable simple ETo methods in Cape Coast by comparing estimated ETo values of three indirect-measurement methods: Hamon, Hargreaves, Blaney Criddle; and two direct-measurement methods: Class A pan and Piche evaporimeter with the Food and Agriculture Organization (FAO) Penman-Monteith equation (FAO56-PM) estimated ETo values. All the methods underestimated ETo values obtained by the FAO56-PM method. However, the estimated ETo values by the Hamon, Hargreaves, and Blaney-Criddle are strongly correlated (R = 0.89, 0.87 and 0.81 respectively), while the Class A pan and Piche evaporimeter methods are weakly correlated (R= 0.37 and 0.31, respectively) with the FAO56-PM method. All but the Piche evaporimeter methods appeared suitable for estimating ETo in the study area. The Class A pan, though weakly correlated with the FAO56-PM method, was also suitable because it had the least mean absolute error (MAE; 0.26 mm day-1) and mean absolute percentage error (MAPE; 6.5%) among the other methods and its ETo curve was closer to the FAO56-PM’s. Key words: Reference crop evapotranspiration, Cape Coast, Food and Agriculture Organization (FAO) Penman-Monteith equation, direct-measurement method.

PyDREAM: high-dimensional parameter inference for biological models in python.

SummaryBiological models contain many parameters whose values are difficult to measure directly via experimentation and therefore require calibration against experimental data. Markov chain Monte Carlo (MCMC) methods are suitable to estimate multivariate posterior model parameter distributions, but these methods may exhibit slow or premature convergence in high-dimensional search spaces. Here, we present PyDREAM, a Python implementation of the (Multiple-Try) Differential Evolution Adaptive Metropolis [DREAM(ZS)] algorithm developed by Vrugt and ter Braak (2008) and Laloy and Vrugt (2012). PyDREAM achieves excellent performance for complex, parameter-rich models and takes full advantage of distributed computing resources, facilitating parameter inference and uncertainty estimation of CPU-intensive biological models.Availability and implementationPyDREAM is freely available under the GNU GPLv3 license from the Lopez lab GitHub repository at http://github.com/LoLab-VU/PyDREAM.Supplementary information Supplementary data are available at Bioinformatics online.

Accelerating Bayesian inference for evolutionary biology models.

MotivationBayesian inference is widely used nowadays and relies largely on Markov chain Monte Carlo (MCMC) methods. Evolutionary biology has greatly benefited from the developments of MCMC methods, but the design of more complex and realistic models and the ever growing availability of novel data is pushing the limits of the current use of these methods.ResultsWe present a parallel Metropolis-Hastings (M-H) framework built with a novel combination of enhancements aimed towards parameter-rich and complex models. We show on a parameter-rich macroevolutionary model increases of the sampling speed up to 35 times with 32 processors when compared to a sequential M-H process. More importantly, our framework achieves up to a twentyfold faster convergence to estimate the posterior probability of phylogenetic trees using 32 processors when compared to the well-known software MrBayes for Bayesian inference of phylogenetic trees.Availability and Implementation https://bitbucket.org/XavMeyer/hogan Supplementary information Supplementary data are available at Bioinformatics online.

On the importance of the long-term seasonal component in day-ahead electricity price forecasting

In day-ahead electricity price forecasting (EPF) the daily and weekly seasonalities are always taken into account, but the long-term seasonal component (LTSC) is believed to add unnecessary complexity to the already parameter-rich models and is generally ignored. Conducting an extensive empirical study involving state-of-the-art time series models we show that (i) decomposing a series of electricity prices into a LTSC and a stochastic component, (ii) modeling them independently and (iii) combining their forecasts can bring – contrary to a common belief – an accuracy gain compared to an approach in which a given time series model is calibrated to the prices themselves.

Decoupled evolution of floral traits and climatic preferences in a clade of Neotropical Gesneriaceae.

BackgroundMajor factors influencing the phenotypic diversity of a lineage can be recognized by characterizing the extent and mode of trait evolution between related species. Here, we compared the evolutionary dynamics of traits associated with floral morphology and climatic preferences in a clade composed of the genera Codonanthopsis, Codonanthe and Nematanthus (Gesneriaceae). To test the mode and specific components that lead to phenotypic diversity in this group, we performed a Bayesian phylogenetic analysis of combined nuclear and plastid DNA sequences and modeled the evolution of quantitative traits related to flower shape and size and to climatic preferences. We propose an alternative approach to display graphically the complex dynamics of trait evolution along a phylogenetic tree using a wide range of evolutionary scenarios.ResultsOur results demonstrated heterogeneous trait evolution. Floral shapes displaced into separate regimes selected by the different pollinator types (hummingbirds versus insects), while floral size underwent a clade-specific evolution. Rates of evolution were higher for the clade that is hummingbird pollinated and experienced flower resupination, compared with species pollinated by bees, suggesting a relevant role of plant-pollinator interactions in lowland rainforest. The evolution of temperature preferences is best explained by a model with distinct selective regimes between the Brazilian Atlantic Forest and the other biomes, whereas differentiation along the precipitation axis was characterized by higher rates, compared with temperature, and no regime or clade-specific patterns.ConclusionsOur study shows different selective regimes and clade-specific patterns in the evolution of morphological and climatic components during the diversification of Neotropical species. Our new graphical visualization tool allows the representation of trait trajectories under parameter-rich models, thus contributing to a better understanding of complex evolutionary dynamics.Electronic supplementary materialThe online version of this article (doi:10.1186/s12862-015-0527-6) contains supplementary material, which is available to authorized users.