Comparison of Outlier-Tolerant Models for Measuring Visual Complexity

Carballal, Adrián; Fernández-Lozano, Carlos; Rodríguez-Fernández, Nereida; Santos, Iria; Romero, Juan

doi:10.3390/e22040488

Cited by 6 publications

(3 citation statements)

References 61 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Although this paper focused only on the potential of greedy cross validation as an accelerant for ML hyperparameter optimization and model selection, the greedy method may prove to be of great value in other situations, as well. There are, for example, many scenarios across a wide range of scientific disciplines in which cross validation is used as a basis for comparing competing models (e.g., [41][42][43]), and the greedy method may be usefully applied to any of these scenarios. Outside of the realm of model selection, the greedy cross validation method developed in this paper may also be usefully applied to a wide variety of bandit problems.…”

Section: Discussionmentioning

confidence: 99%

Greed Is Good: Rapid Hyperparameter Optimization and Model Selection Using Greedy k-Fold Cross Validation

Soper

2021

Electronics

View full text Add to dashboard Cite

Selecting a final machine learning (ML) model typically occurs after a process of hyperparameter optimization in which many candidate models with varying structural properties and algorithmic settings are evaluated and compared. Evaluating each candidate model commonly relies on k-fold cross validation, wherein the data are randomly subdivided into k folds, with each fold being iteratively used as a validation set for a model that has been trained using the remaining folds. While many research studies have sought to accelerate ML model selection by applying metaheuristic and other search methods to the hyperparameter space, no consideration has been given to the k-fold cross validation process itself as a means of rapidly identifying the best-performing model. The current study rectifies this oversight by introducing a greedy k-fold cross validation method and demonstrating that greedy k-fold cross validation can vastly reduce the average time required to identify the best-performing model when given a fixed computational budget and a set of candidate models. This improved search time is shown to hold across a variety of ML algorithms and real-world datasets. For scenarios without a computational budget, this paper also introduces an early stopping algorithm based on the greedy cross validation method. The greedy early stopping method is shown to outperform a competing, state-of-the-art early stopping method both in terms of search time and the quality of the ML models selected by the algorithm. Since hyperparameter optimization is among the most time-consuming, computationally intensive, and monetarily expensive tasks in the broader process of developing ML-based solutions, the ability to rapidly identify optimal machine learning models using greedy cross validation has obvious and substantial benefits to organizations and researchers alike.

show abstract

Section: Discussionmentioning

confidence: 99%

Greed Is Good: Rapid Hyperparameter Optimization and Model Selection Using Greedy k-Fold Cross Validation

Soper

2021

Electronics

View full text Add to dashboard Cite

show abstract

“…However, this will only be possible when raster land use data that includes OA as a class is integrated into 2.5D analysis. Theoretically, VP could also be investigated with the use of visual complexity theory, which uses regularity and heterogeneity of a number of elements to measure a perceived complexity value and reference it to image evaluation by a group of humans [88]. Visual complexity theory has so far been applied in art object aesthetic evaluation [89] as well as advertising content design analysis [90].…”

Section: Discussionmentioning

confidence: 99%

Chaos in Motion: Measuring Visual Pollution with Tangential View Landscape Metrics

Chmielewski

2020

Land

View full text Add to dashboard Cite

Visual pollution (VP) in the form of outdoor advertisements (OA) is a threat to landscape physiognomy. Despite their proven usefulness in landscape aesthetic studies, landscape metrics have not yet been applied to address the phenomenon of VP. To fill this knowledge gap, a methodological framework for the measurement of VP using tangential view landscape metrics is proposed, which is accompanied by statistically significant proofs. Raster products derived from aerial laser scanning data were used to characterize two study areas with different topographic conditions in the city of Lublin, East Poland. The visibility of the cityscape in motion was simulated through viewshed measurements taken at equal intervals in the forwards and backwards directions along pedestrian walkways. The scrutinized tangential view landscape metrics (visible area, maximum visible distance, skyline, Shannon depth, view depth line) was the object of a two-fold interpretation wherein the spatial occurrence of VP as well as its impacts on the visual landscape character (VLC) were examined. The visible area metrics were found to be highly sensitive VP indicators. The maximum visible distance metrics provided evidence for the destructive effect of OA on view corridors. The Shannon depth and depth line metrics were not found to be statistically significant indicators of VP. Results from directional viewshed modelling indicate that distortion in the analyzed cityscape physiognomy depends on the view direction. The findings allow for particular recommendations with practical implementations in land use planning, which are discussed along with limitations to our proposed methods.

show abstract

“…However, this method introduces human bias [ 15 ] when the intended receivers of complex stimuli are non-human animals. More recently, however, other methods of quantifying complexity have been developed, which have aimed to take into account the sensory system and psychology of receivers [ 1 , 2 , 13 , 16 ]. For measures of visual complexity to be biologically relevant, they must be based on receiver perception.…”

Section: Introductionmentioning

confidence: 99%

Visual complexity of egg patterns predicts egg rejection according to Weber's law

et al. 2022

View full text Add to dashboard Cite

Visual complexity is ubiquitous in nature. Drivers of complexity include selection in coevolutionary arms races between antagonists. However, the causes and consequences of biological complexity and its perception are largely understudied, partly because complexity is difficult to quantify. Here, we address this by studying egg pattern complexity and its perception in hosts (tawny-flanked prinia Prinia subflava ), which visually recognize and reject mimetic eggs of their virulent brood parasite (cuckoo finch Anomalospiza imberbis ). Using field data and an optimization algorithm, we compute a complexity metric which predicts rejection of experimentally placed conspecific eggs in prinia nests. Real cuckoo finch eggs exhibit significantly lower pattern complexity than prinia eggs, suggesting that high complexity benefits hosts because it distinguishes host eggs from parasitic eggs. We show that prinias perceive complexity differences according to Weber's law of proportional processing (i.e. relative, rather than absolute, differences between stimuli are processed in discrimination, such that two eggs with simple patterns are more easily discriminable than two with complex patterns). This may influence coevolutionary trajectories of hosts and parasites. The new methods presented for quantifying complexity and its perception can help us to understand selection pressures driving the evolution of complexity and its consequences for species interactions.

show abstract

Comparison of Outlier-Tolerant Models for Measuring Visual Complexity

Cited by 6 publications

References 61 publications

Greed Is Good: Rapid Hyperparameter Optimization and Model Selection Using Greedy k-Fold Cross Validation

Greed Is Good: Rapid Hyperparameter Optimization and Model Selection Using Greedy k-Fold Cross Validation

Chaos in Motion: Measuring Visual Pollution with Tangential View Landscape Metrics

Visual complexity of egg patterns predicts egg rejection according to Weber's law

Contact Info

Product

Resources

About