Automatic evolutionary learning of composite models with knowledge enrichment

Kalyuzhnaya, Anna V.; Nikitin, Nikolay O.; Vychuzhanin, Pavel; Hvatov, Alexander; Boukhanovsky, Alexander V.

doi:10.1145/3377929.3398167

Cited by 13 publications

(13 citation statements)

References 2 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For the model’s generative design task, the most expensive step usually refers to the evaluation of the fitness function value [ 6 ]. The calculation of the fitness function for the individuals of the evolutionary algorithm can be parallelized in different ways that are presented in Figure 5 .…”

Section: Important Obstacles On the Way Of Generative Co-design Immentioning

confidence: 99%

“…The framework allows generating composite models using evolutionary approaches. The composite model generated by the framework can include different types of models [ 6 ]. The following parameters of the genetic algorithm were used during the experiments: maximum number of the generations in 20, number of the individuals in each population is 32, probability of mutation, probability of mutation is 0.8, probability of crossover is 0.8, maximum arity of the composite model is 4, maximum depth of the composite model is 3.…”

Section: Experimental Studiesmentioning

confidence: 99%

“…A composite model has a heterogeneous structure, so it can include models of different nature: machine learning (ML), equation-based, etc. [ 6 ].…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Towards Generative Design of Computationally Efficient Mathematical Models with Evolutionary Learning

Kalyuzhnaya

Nikitin

Hvatov

et al. 2020

Entropy

Self Cite

View full text Add to dashboard Cite

In this paper, we describe the concept of generative design approach applied to the automated evolutionary learning of mathematical models in a computationally efficient way. To formalize the problems of models’ design and co-design, the generalized formulation of the modeling workflow is proposed. A parallelized evolutionary learning approach for the identification of model structure is described for the equation-based model and composite machine learning models. Moreover, the involvement of the performance models in the design process is analyzed. A set of experiments with various models and computational resources is conducted to verify different aspects of the proposed approach.

show abstract

Section: Important Obstacles On the Way Of Generative Co-design Immentioning

confidence: 99%

Section: Experimental Studiesmentioning

confidence: 99%

See 1 more Smart Citation

Towards Generative Design of Computationally Efficient Mathematical Models with Evolutionary Learning

Kalyuzhnaya

Nikitin

Hvatov

et al. 2020

Entropy

Self Cite

View full text Add to dashboard Cite

show abstract

“…However, most of the existing AutoML tools cannot be used to obtain heterogeneous (composite) pipelines, that combined models of different nature (for example, ML models and hydrological models). At the same time, composite modeling is especially promising for flood forecasting since it allows combining the state-of-the-art data-driven and physicsbased methods and models according to the Composite Artificial Intelligence concepts [17].…”

Section: Introduction 1modelling Of the River Floodsmentioning

confidence: 99%

Short-Term River Flood Forecasting Using Composite Models and Automated Machine Learning: The Case Study of Lena River

Sarafanov

Borisova

Maslyaev

et al. 2021

Water

Self Cite

View full text Add to dashboard Cite

The paper presents a hybrid approach for short-term river flood forecasting. It is based on multi-modal data fusion from different sources (weather stations, water height sensors, remote sensing data). To improve the forecasting efficiency, the machine learning methods and the Snowmelt-Runoff physical model are combined in a composite modeling pipeline using automated machine learning techniques. The novelty of the study is based on the application of automated machine learning to identify the individual blocks of a composite pipeline without involving an expert. It makes it possible to adapt the approach to various river basins and different types of floods. Lena River basin was used as a case study since its modeling during spring high water is complicated by the high probability of ice-jam flooding events. Experimental comparison with the existing methods confirms that the proposed approach reduces the error at each analyzed level gauging station. The value of Nash–Sutcliffe model efficiency coefficient for the ten stations chosen for comparison is 0.80. The other approaches based on statistical and physical models could not surpass the threshold of 0.74. Validation for a high-water period also confirms that a composite pipeline designed using automated machine learning is much more efficient than stand-alone models.

show abstract

“…In recent years, machine learning and deep learning techniques obtain a complex network by training a rich set of the known data to identify the relationships between the independent and dependent parameters. These algorithms reduce the inter-collinearity of adjacent pixels in the above-mentioned models in both spatial and temporal domains, and have great advantages in solving nonlinear problems [24][25][26][27][28]. However, the need of an intensive training dataset and high computational capacity may hinder their widespread application.…”

Section: Introductionmentioning

confidence: 99%

Gap-Filling of 8-Day Terra MODIS Daytime Land Surface Temperature in High-Latitude Cold Region with Generalized Additive Models (GAM)

et al. 2021

View full text Add to dashboard Cite

Land surface temperature (LST) is a crucial parameter driving the dynamics of the thermal state on land surface. In high-latitude cold region, a long-term, stable LST product is of great importance in examining the distribution and degradation of permafrost under pressure of global warming. In this study, a generalized additive model (GAM) approach was developed to fill the missing pixels of the MODIS/Terra 8-day Land Surface Temperature (MODIS LST) daytime products with the ERA5 Land Skin Temperature (ERA5ST) dataset in a high-latitude watershed in Eurasia. Comparison at valid pixels revealed that the MODIS LST was 4.8–13.0 °C higher than ERA5ST, which varies with land covers and seasons. The GAM models fairly explained the LST differences between the two products from multiple covariates including satellite-extracted environmental variables (i.e., normalized difference water index (NDWI), normalized difference vegetation index (NDVI), and normalized difference snow index (NDSI) as well as locational information. Considering the dramatic seasonal variation of vegetation and frequent snow in the cold region, the gap-filling was conducted in two seasons. The results revealed the root mean square errors (RMSE) of 2.7 °C and 3.4 °C between the valid MODIS LST and GAM-simulated LST data in the growing season and snowing season, respectively. By including the satellite-extracted land surface information in the GAM model, localized variations of land surface temperature that are often lost in the reanalysis data were effectively compensated. Specifically, land surface wetness (NDWI) was found to be the greatest contributor to explaining the differences between the two products. Vegetation (NDVI) was useful in the growing season and snow cover (NDSI) cannot be ignored in the snow season of the study region. The km-scale gap-filled MODIS LST products provide spatially and temporally continuous details that are useful for monitoring permafrost degradation in cold regions in scenarios of global climate change.

show abstract

Automatic evolutionary learning of composite models with knowledge enrichment

Cited by 13 publications

References 2 publications

Towards Generative Design of Computationally Efficient Mathematical Models with Evolutionary Learning

Towards Generative Design of Computationally Efficient Mathematical Models with Evolutionary Learning

Short-Term River Flood Forecasting Using Composite Models and Automated Machine Learning: The Case Study of Lena River

Gap-Filling of 8-Day Terra MODIS Daytime Land Surface Temperature in High-Latitude Cold Region with Generalized Additive Models (GAM)

Contact Info

Product

Resources

About