Search research articles

ABOUT JoVE

Overview Leadership Blog JoVE Help Center

AUTHORS

Publishing Process Editorial Board Scope & Policies Peer Review FAQ Submit

LIBRARIANS

Testimonials Subscriptions Access Resources Library Advisory Board FAQ

RESEARCH

JoVE Journal Methods Collections JoVE Encyclopedia of Experiments Archive

EDUCATION

JoVE Core JoVE Business JoVE Science Education JoVE Lab Manual Faculty Resource Center Faculty Site

Terms & Conditions of Use

Search research articles

Related Experiment Videos

Learning Tetris using the noisy cross-entropy method.

István Szita¹, András Lörincz

¹szityu@eotvos.elte.hu

Neural Computation

|October 21, 2006

Summary

This summary is machine-generated.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by

Same author

Incidence, disease course, therapeutic strategies and outcomes of inflammatory bowel disease-unclassified patients in Western Hungary: A population-based cohort.

World journal of gastroenterology·2025

Same author

Inflammatory bowel diseases in the elderly population: epidemiology, long-term disease course, surgery rates, and biological use-data from the Veszprem county cohort between 1977 and 2020.

Journal of Crohn's & colitis·2025

Same author

Burden of perianal disease in Crohn's disease: Accelerating medical therapy and high rates of perianal surgery over the last four decades - Results from a population-based study over four decades.

Alimentary pharmacology & therapeutics·2023

Same author

BlinkLinMulT: Transformer-Based Eye Blink Detection.

Journal of imaging·2023

Same author

Declining Trends of Reoperations and Disease Behaviour Progression in Crohn's Disease over Different Therapeutic Eras-A Prospective, Population-Based Study from Western Hungary between 1977-2020, Data from the Veszprem Cohort.

Journal of Crohn's & colitis·2023

Same author

Cross-Viewpoint Semantic Mapping: Integrating Human and Robot Perspectives for Improved 3D Semantic Reconstruction.

Sensors (Basel, Switzerland)·2023

Noise injection enhances the cross-entropy method in reinforcement learning. This approach prevents premature convergence, leading to superior policies in complex tasks like Tetris.

Area of Science:

Artificial Intelligence
Machine Learning
Computational Neuroscience

Background:

The cross-entropy method is a powerful optimization technique.
Its application in reinforcement learning (RL) is hindered by convergence to suboptimal policies.

Purpose of the Study:

To improve the cross-entropy method's performance in reinforcement learning.
To address the issue of premature convergence to suboptimal policies.

Main Methods:

Applying noise injection to the cross-entropy method.
Utilizing the game of Tetris as an experimental environment for demonstration.

Main Results:

The modified cross-entropy method successfully prevented early convergence.

Related Experiment Videos

The resulting reinforcement learning policy demonstrated a significant performance improvement, outperforming prior algorithms by nearly two orders of magnitude.

Conclusions:

Noise injection is an effective strategy for enhancing the cross-entropy method in reinforcement learning.
This technique offers a promising direction for developing more efficient and effective RL algorithms.