Effect Of Model Complexity Research Articles

The effect of model complexity on the human center of mass estimation using the statically equivalent serial chain technique

Estimating the human center of mass (CoM) has long been recognized as a highly complex process. A relatively recent and noteworthy technique for CoM estimation that has gained popularity is the statically equivalent serial chain (SESC). This technique employs a remodeling of the human skeleton as a serial chain where the end effector represents the CoM location. In this study, we aimed to evaluate the impact of model complexity on the estimation capability of the SESC technique. To achieve this, we designed and rigorously assessed four distinct models with varying complexities against the static center of pressure (CoP) as reference, by quantifying both the root-mean-square (RMS) and correlation metrics. In addition, the Bland–Altman analysis was utilized to quantify the agreement between the estimations and reference values. The findings revealed that increasing the model complexity significantly improved CoM estimation quality up to a specific threshold. The maximum observed RMS difference among the models reached 9.85 mm. However, the application and task context should be considered, as less complex models still provided satisfactory estimation performance. In conclusion, the evaluation of model complexity demonstrated its impact on CoM estimation using the SESC technique, providing insights into the trade-off between accuracy and complexity in practical applications.

Data assimilation of physical and chlorophyll a observations in the California Current System using two biogeochemical models

Biogeochemical numerical models coupled to physical ocean circulation models are commonly combined with data assimilation in order to improve the models’ state or parameter estimates. Yet much still needs to be learned about important aspects of biogeochemical data assimilation, such as the effect of model complexity and the importance of more realistic model formulations on assimilation results. In this study, 4D-Var-based state estimation is applied to two biogeochemical ocean models: a simple NPZD model with 4 biogeochemical variables (including 1 phytoplankton, 1 zooplankton) and the more complex NEMURO model, containing 11 biogeochemical variables (including 2 phytoplankton, 3 zooplankton). Both models are coupled to a 3-dimensional physical ocean circulation model of the U.S. west coast based on the Regional Ocean Modelling System (ROMS). Chlorophyll satellite observations and physical observations are assimilated into the model, yielding substantial improvements in state estimates for the observed physical and biogeochemical variables in both model formulations. In comparison to the simpler NPZD model, NEMURO shows a better overall fit to the observations. The assimilation also results in small improvements for simulated nitrate concentrations in both models and no apparent degradation of the output for other unobserved variables. The forecasting skill of the biogeochemical models is strongly linked to model performance without data assimilation: for both models, the improved fit obtained through assimilation degrades at similar relative rates, but drops to different absolute levels. Despite the better performance of NEMURO in our experiments, the choice of model and desired level of complexity should depend on the model application and the data available for assimilation.

The effect of model complexity in simulating unsaturated zone flow processes on recharge estimation at varying time scales

Summary Recent increases in computational power have led to the development of more advanced physically-based models which can handle a wide range of environmental processes. Although these models are very useful for increasing our understanding of unsaturated zone flow processes, their outputs usually contain high uncertainty, particularly when the level of complexity is not supported by observations. In this context, the aim of this paper is to compare the performance of three different model conceptualizations of a shallow unsaturated soil zone using the physically-based model HydroGeoSphere (HGS). To accomplish this task, we simulated actual evapotranspiration (ET), water content (WC) and discharge (D) from a weighing lysimeter for each of the conceptual models. Conceptual Model 1 considers the lysimeter as a homogeneous zone with matrix flow, while Conceptual Model 2 has an added preferential flow component. Conceptual Model 3 includes layered heterogeneity in addition to the matrix and preferential flow components. The results indicated that the model performance in reproducing daily ET, WC and D improves when we move from simple models to more complex models. A comparison between event-based, monthly, seasonal and yearly time scales indicates that the simplest conceptual model is not reliable for reproducing event-based discharges. However, it can compete with more complex models at annual scales, although the uncertainty bound for the simple model is very high. While increasing complexity from the simplest to the more complex model leads to lower uncertainty bounds and more reliable values of the lysimeter discharge at monthly and seasonal time scales, uncertainty bounds became larger when complexity increased in the most complex model. This is related to a higher number of unknown model parameters in the calibration which are not supported by the available observation datasets.

Data fusion with Gaussian processes

This paper addresses the problem of fusing multiple sets of heterogeneous sensor data using Gaussian processes (GPs). Experiments on large scale terrain modeling in mining automation are presented. Three techniques in increasing order of model complexity are discussed. The first is based on adding data to an existing GP model. The second approach treats data from different sources as different noisy samples of a common underlying terrain and fusion is performed using heteroscedastic GPs. The final approach, based on dependent GPs, models each data set by a separate GP and learns spatial correlations between data sets through auto and cross covariances. The paper presents a unifying view of approaches to data fusion using GPs, a statistical evaluation that compares these approaches and multiple previously untested variants of them and an insight into the effect of model complexity on data fusion. Experiments suggest that in situations where data being fused is not rich enough to require a complex GP data fusion model or when computational resources are limited, the use of simpler GP data fusion techniques, which are constrained versions of the more generic models, reduces optimization complexity and consequently can enable superior learning of hyperparameters, resulting in a performance gain.

Finite element analysis of aircraft tyre – Effect of model complexity on tyre performance characteristics

A brief study on the complexity of the structure of aircraft tyres is presented. The main goal is to investigate how complexity of Finite element (FE) models of the tyre can affect the simulation behaviour with respect to accurate prediction of real tyre behaviour. Three FE models, so called “simple”, “regular” and “complex” were developed for this investigation so that the accuracy of results obtained from the models for different types of analysis can be determined in comparison with measured test data. For all the simulations carried out, including tyre burst test and tyre inflation, a hyperelastic material property was assumed for modelling rubber and cord materials, using the Yeoh model [1]. Simulations were performed using Abaqus/CAE for two dimensional (2D) analysis and Abaqus command line partly for three dimensional (3D) simulation of tyre inflation. Physical tests were carried out on two tyres for FE model validation in terms of tyre burst test and profile sizing measurements. The results showed excellent accuracy in terms of deformation in FE models by comparison with real size measurements of tyre profiles. Also, a quite good prediction of tyre burst pressure under high inflation was obtained for the different models. Finally, the investigation showed that while maximum stress in cords and maximum deflection of tyres at the rated inflation pressure were fairly insensitive to varying the mesh size in the FE models, maximum stress in the rubber components was particularly sensitive to such mesh size variation.

ICHASM, a flexible land-surface model that incorporates stable water isotopes

Isotope-enabled Global Climate Models (GCMs) have been previously used to investigate water isotope fluxes in present and past climate systems. Water isotope modelling is also important in modern ecosystem studies, for investigating the source and transport of moisture and carbon dioxide fluxes. However, in this modelling, there has only been limited investigation of the effect of parameterisation complexity on stable water isotope partitioning. The present study has two aims. The first is to incorporate a stable water isotope parameterisation into a flexible land surface scheme (LSS): the CHAmeleon Surface Model (CHASM). This scheme offers five different modes, each containing up to four water reservoirs (canopy interception, snowpack, root zone and bare ground storage), and each mode offers the opportunity to modify resistances to evaporative fluxes from these reservoirs. These different modes allow the effect of model complexity on different parameterisations to be explored within a common modelling framework. Two modelling experiments are reported which use the iPILPS Phase 1 forcing: the ‘control’ experiment and one using a simpler mode of CHASM comprising a bucket hydrology plus a fixed stomatal resistance. The additional set of experiments using iCHASM's variable modes shows, contrary to an earlier experiments with CHASM, that a land surface model with only a bucket hydrology scheme and constant surface resistance cannot reproduce the behaviour of an LSS that has additional functionality (such as bare ground evaporation, and canopy interception and aerodynamic, surface and stomatal resistances).

SEMI-SUPERVISED SEQUENCE CLASSIFICATION WITH HMMs

Using unlabeled data to help supervised learning has become an increasingly attractive methodology and proven to be effective in many applications. This paper applies semi-supervised classification algorithms, based on hidden Markov models, to classify sequences. For model-based classification, semi-supervised learning amounts to using both labeled and unlabeled data to train model parameters. We examine three different strategies of using labeled and unlabeled data in the model training process. These strategies differ in how and when labeled and unlabeled data contribute to the model training process. We also compare regular semi-supervised learning, where there are separate unlabeled training data and unlabeled test data, with transductive learning where we do not differentiate between unlabeled training data and unlabeled test data. Our experimental results on synthetic and real EEG time-series show that substantially improved classification accuracy can be achieved by these semi-supervised learning strategies. The effect of model complexity on semi-supervised learning is also studied in our experiments.

Effect of model complexity and gait criteria on the synthesis of bipedal locomotion.

The control functions, system motions, and other characteristics which are obtained from the mathematical synthesis of ``optimal'' bipedal locomotion gaits with respect to locomotion criteria (to be minimized) defined in the paper are considered. Results are presented for three mathematical biped models and several locomotion criteria (objective functions). Optimal walking trajectories and associated control functions are synthesized using the optimal programming method of [1]. In the synthesis approach, the model structure and locomotion criterion determine the optimal controls motion and foot placement parameters. These effects are examined, and qualitative comparisons to human walking patterns are suggested. These studies also provide quantitative information for the design of prostheses and walking machines, for example: ranges of suitable motion patterns (as functions of time) and magnitudes of controlling efforts and energy requirements.