Structural Evolutionary Learning for Composite Classification Models

Nikitin, Nikolay O.; Polonskaia, Iana S.; Vychuzhanin, Pavel; Barabanova, Irina V.; Kalyuzhnaya, Anna V.

doi:10.1016/j.procs.2020.11.043

Cited by 14 publications

(9 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The models below the threshold are removed from the tree iteratively with metric recalculation after each removal. Also, the applied implementation of the regularization operator can be task-specific (e.g., custom regularization for composite machine learning models [11] and the LASSO regression for the partial differential equations [9]). Unlike the genetic operators defined in general, the objectives are defined for the given class of the atomic models and the given problem.…”

Section: Problem Statement For Model-agnostic Approachmentioning

confidence: 99%

“…Quality and robustness could be considered as an example of the model's properties. The proposed model-agnostic approach can be used to discover the robust composite machine learning models with the structure described as a directed acyclic graph (as described in [11]). In this case, the building blocks are regression-based machine learning models, algorithms for feature selection, and feature transformation.…”

Section: Composite Machine Learning Modelsmentioning

confidence: 99%

See 1 more Smart Citation

Model-agnostic multi-objective approach for the evolutionary discovery of mathematical models

Hvatov¹,

Maslyaev²,

Polonskaya³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

In modern data science, it is often not enough to obtain only a data-driven model with a good prediction quality. On the contrary, it is more interesting to understand the properties of the model, which parts could be replaced to obtain better results. Such questions are unified under machine learning interpretability questions, which could be considered one of the area's raising topics. In the paper, we use multi-objective evolutionary optimization for composite data-driven model learning to obtain the algorithm's desired properties. It means that whereas one of the apparent objectives is precision, the other could be chosen as the complexity of the model, robustness, and many others. The method application is shown on examples of multi-objective learning of composite models, differential equations, and closed-form algebraic expressions are unified and form approach for model-agnostic learning of the interpretable models.

show abstract

Section: Problem Statement For Model-agnostic Approachmentioning

confidence: 99%

Section: Composite Machine Learning Modelsmentioning

confidence: 99%

Model-agnostic multi-objective approach for the evolutionary discovery of mathematical models

Hvatov¹,

Maslyaev²,

Polonskaya³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…Such scheme performs the dynamic adaptation of operators' probabilistic rates on the level of the population ( [49]). In [50] examples of evolutionary operators for the graph structures (crossovers, mutations, and regularization) with a detailed description.…”

Section: Automated Evolutionary Designmentioning

confidence: 99%

Automated Evolutionary Approach for the Design of Composite Machine Learning Pipelines

Nikitin,

Vychuzhanin,

Sarafanov

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

The effectiveness of the machine learning methods for real-world tasks depends on the proper structure of the modeling pipeline. The proposed approach is aimed to automate the design of composite machine learning pipelines, which is equivalent to computation workflows that consist of models and data operations. The approach combines key ideas of both automated machine learning and workflow management systems. It designs the pipelines with a customizable graph-based structure, analyzes the obtained results, and reproduces them. The evolutionary approach is used for the flexible identification of pipeline structure. The additional algorithms for sensitivity analysis, atomization, and hyperparameter tuning are implemented to improve the effectiveness of the approach. Also, the software implementation on this approach is presented as an open-source framework. The set of experiments is conducted for the different datasets and tasks (classification, regression, time series forecasting). The obtained results confirm the correctness and effectiveness of the proposed approach in the comparison with the state-of-the-art competitors and baseline solutions.

show abstract

“…None of these are good ways when we need the highly automated tool, that can be adapted to different properties of input time series. That is why, secondly, we suggest learning forward and backward models M F and M B as composite models [20] fitted on quasi-local data. And, third, we suggest joining predictions with forward and backward models by another learned ensemble model Ω (a kind of stacking approach).…”

Section: Problem Statementmentioning

confidence: 99%

“…It allows building the data-driven and hybrid models consist of several atomic blocks. For the time series forecasting, the chains consist of several regression models that can be generated using evolutionary structural learning [20]. It should be noted that alternative AutoML frameworks exist (e.g.…”

Section: Fedot Frameworkmentioning

confidence: 99%

Automated data-driven approach for gap filling in the time series using evolutionary learning

Sarafanov¹,

Nikitin²,

Kalyuzhnaya³

2021

Preprint

Self Cite

View full text Add to dashboard Cite

Time series analysis is widely used in various fields of science and industry. However, the vast majority of the time series obtained from real sources contain a large number of gaps, have a complex character, and can contain incorrect or missed parts. So, it is useful to have a convenient, efficient, and flexible instrument to fill the gaps in the time series. In this paper, we propose an approach for filling the gaps by the evolutionary automatic machine learning, that is implemented as a part of the FEDOT framework. Automated identification of the optimal datadriven model structure allows the adopting of the gap filling strategy to the specific problem. As a case study, the multivariate sea surface height dataset is used. During the experimental studies, the proposed approach was compared with other gap-filling methods and the composite models allow obtaining the higher quality of the gap restoration.

show abstract

Structural Evolutionary Learning for Composite Classification Models

Cited by 14 publications

References 20 publications

Model-agnostic multi-objective approach for the evolutionary discovery of mathematical models

Model-agnostic multi-objective approach for the evolutionary discovery of mathematical models

Automated Evolutionary Approach for the Design of Composite Machine Learning Pipelines

Automated data-driven approach for gap filling in the time series using evolutionary learning

Contact Info

Product

Resources

About