A Genetic Algorithm-Based, Hybrid Machine Learning Approach to Model Selection

@article{Bies2006AGA,
  title={A Genetic Algorithm-Based, Hybrid Machine Learning Approach to Model Selection},
  author={Robert R Bies and Matthew F. Muldoon and Bruce G. Pollock and Stephen B. Manuck and Gwenn S. Smith and Mark E. Sale},
  journal={Journal of Pharmacokinetics and Pharmacodynamics},
  year={2006},
  volume={33},
  pages={195-221}
}
We describe a general and robust method for identification of an optimal non-linear mixed effects model. This includes structural, inter-individual random effects, covariate effects and residual error models using machine learning. This method is based on combinatorial optimization using genetic algorithm. 

Generalized Net model of selection operator of genetic algorithms

The apparatus of Generalized Nets is applied here to a description of a selection operator, which is one of the basic genetic algorithm operators. This genetic operator performs a probabilistic

A genetic algorithm based global search strategy for population pharmacokinetic/pharmacodynamic model selection

TLDR
These methods are reviewed, with emphasis on genetic algorithm approaches, and the role these methods may play in population pharmacokinetic/pharmacodynamic model selection is discussed.

Application of a single-objective, hybrid genetic algorithm approach to pharmacokinetic model building

TLDR
The single-objective, hybrid genetic algorithm represents a general pharmacokinetics model building methodology whose ability to rapidly search the feasible solution space leads to nearly equivalent or superior model fits to pharmacokinetic data.

Population pharmacokinetic model selection assisted by machine learning

TLDR
This manuscript compared the classical pharmacometric approach with two machine learning methods, genetic algorithm and neural networks, in different scenarios based on simulated pharmacokinetic data, and suggested that machine learning approaches can achieve a first fast selection of models which can be followed by more conventional pharmacometric approaches.

Generalized Net Models of Basic Genetic Algorithm Operators

Generalized nets (GN) are applied here to describe some basic operators of genetic algorithms, namely selection, crossover and mutation and different functions for selection (roulette wheel selection

Developing an adaptation process for real-coded genetic algorithms

TLDR
Experimental results show that this new process accelerated the algorithm and a certain solution has been reached in fewer generations, and better solutions were achieved, especially for a certain number of generations.

Identification of Optimal Kernel Parameters of RKHS model based on Genetics Algorithm

This paper proposes a new approach based on the Genetics Algorithm to determine the optimal kernel parameters of the Reproducing Kernel Hilbert Space (RKHS) model. These parameters are the width of

Probability Estimation by an Adapted Genetic Algorithm in Web Insurance

TLDR
This work proposes to use a genetic algorithm specifically adapted to overcome the statistical method defaults and shows its performances on real datasets provided by the company MeilleureAssurance.com.

Modelling of a Roulette Wheel Selection Operator in Genetic Algorithms Using Generalized Nets

The apparatus of Generalized Nets (GN) is applied here to a description of a selection operator, which is one of the basic genetic algorithm operators. The GN model presented here describes roulette
...

References

SHOWING 1-10 OF 27 REFERENCES

Hybrid Genetic Algorithms for Feature Selection

TLDR
Experiments revealed that the proposed hybrid GA is superior to both a simple GA and sequential search algorithms, and showed better convergence properties compared to the classical GAs.

Index Tracking : Genetic Algorithms for Investment Portfolio Selection

TLDR
This project was concerned with passive portfolio selection using genetic algorithms and quadratic programming techniques to find a large universal set of shares for a subset that performs well, so a stochastic search method must be used.

Information Theory and an Extension of the Maximum Likelihood Principle

TLDR
The classical maximum likelihood principle can be considered to be a method of asymptotic realization of an optimum estimate with respect to a very general information theoretic criterion to provide answers to many practical problems of statistical model fitting.

Genetic algorithm optimization in building portfolio management

A significant proportion of building investment expenditure goes to replacement expenditure for organizations owning a large building stock or portfolio. Over the years, researchers have attempted to

Model Building for Nonlinear Mixed Effects Models

Nonlinear mixed-effects models involve both fixed effects and random effects. Model building for nonlinear mixed-effects models is the process of determining the characteristics of both the fixed and

Efficient Screening of Covariates in Population Models Using Wald's Approximation to the Likelihood Ratio Test

We propose an efficient algorithm for screening covariates in population model building using Wald's approximation to the likelihood ratio test (LRT) statistic in conjunction with Schwarz's Bayesian

Combinatorial Optimization: Algorithms and Complexity

This clearly written , mathematically rigorous text includes a novel algorithmic exposition of the simplex method and also discusses the Soviet ellipsoid algorithm for linear programming; efficient

Selection of predictor variables for pneumonia using neural networks and genetic algorithms.

TLDR
Variable sets derived using a genetic algorithm for neural networks accurately discriminated pneumonia from other respiratory conditions, and did so with greater accuracy than variables derived using stepwise neural networks or logistic regression in some cases.

Prediction Discrepancies for the Evaluation of Nonlinear Mixed-Effects Models

TLDR
A criterion and a test to evaluate nonlinear mixed-effects models based on the whole predictive distribution and a proposal to compute prediction discrepancies using Monte Carlo integration which does not require model approximation are developed.

Building population pharmacokineticpharmacodynamic models. I. Models for covariate effects

TLDR
A previously described approach to efficiently find the relationships between the PK-PD parameters and covariates is extended and the time required to derive a population model is markedly reduced because the number of necessary NONMEM runs is reduced.