Code Monkey home page Code Monkey logo

roman91de / master_thesis Goto Github PK

View Code? Open in Web Editor NEW
2.0 1.0 0.0 163.36 MB

Master Thesis for M.Sc. Business Education - Pre-Trained Denoising Autoencoders Long Short-Term Memory Networks as probabilistic Models for Estimation of Distribution Genetic Programming

TeX 84.51% Jupyter Notebook 2.27% Mermaid 1.33% Python 11.89%
denoising-autoencoders estimation-distribution-algorithm genetic-programming long-short-term-memory meta-heuristics pre-training symbolic-regression

master_thesis's Introduction

Master Thesis

Topic: Pre-Trained Denoising Autoencoders Long Short-Term Memory Networks as probabilistic Models for Estimation of Distribution Genetic Programming

Institution: Johannes Gutenberg University Mainz, Chair of Business Administration and Computer Science (FB 03)

Abstract

English

Denoising Autoencoder Genetic Programming (DAE-GP) is an Estimation of Distribution Algorithm in the domain of Genetic Programming that uses Denoising Autoencoders Long Short-Term Memory Networks (DAE-LSTM) as probabilistic models for sampling new populations of solutions. This thesis investigates the possible benefits and downsides of using pre-training for the DAE-LSTM networks of DAE-GP for four real world symbolic regression problems. The experiments conducted did show that pre-training can drastically reduce the number of epochs that are necessary for the DAE-LSTM training at each generation of the DAE-GP search. Another interesting finding was that pre-training also increases the levenshtein edit distance between individual solutions inside the population which is a metric for the diversity of a population. Unfortunately, pre-training did not yield any improvements in the final fitness or size of solutions, despite significantly increasing the total run-time for DAE-GP.

Deutsche Fassung

Denoising Autoencoder Genetic Programming (DAE-GP) ist ein Estimation of Distribution Algorithmus (EDA) aus dem Forschungfeld der genetischen Programmierung (GP). In DAE-GP werden Denoising Autoencoders Long Short-Term Memory Netzwerke (DAE-LSTM) als probabilistische Modelle verwendet um neue Populationen von Lösungen für eine evolutionäre Suche zu erzeugen. Diese Masterarbeit untersucht die Vor- und Nachteile des Einsatzes einer Pre-Training Strategie für die DAE-LSTM Netzwerke von DAE-GP an vier Datensätzen für symbolische Regression. Die durchgeführten Experimente haben gezeigt, dass Pre-Training die Anzahl von Trainingsepochen für die DAE-LSTM Netzwerke in jeder Generation statistisch signifikant reduzieren konnte. Außerdem zeigt sich, dass Pre-Training die Levenshtein Editierdistanz, ein Maß für die Populationsdiversität, signifikant erhöhen konnte. Leider konnte Pre-Training die Qualität der jeweils besten gefundenen Lösungen, weder im Bezug auf ihre Fitness noch auf ihre Größe, verbessern. Auch führte Pre-Training in den durchgeführten Experimenten zu einer starken Erhöhung der Laufzeit des DAE-GP Algorithmus.

Keywords

Genetic Programming, Estimation of Distribution Algorithms, Denoising Autoencoder Genetic Programming, Pre-Training, Long Short-Term Memory Networks, Symbolic Regression

Documents available:

Full Thesis: paper.pdf (RMarkdown Source File: paper.Rmd)

PDF-Slide Show: slides.pdf (RMarkdown Source File: slides.Rmd)

master_thesis's People

Contributors

roman91de avatar

Stargazers

 avatar

Watchers

 avatar

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.