Clarification on NumEpoch, MaxMiniBatchPerEpoch, and LearningFrequency in rlAgentDDPGOptions

35 visualizaciones (últimos 30 días)
Fabián
Fabián el 13 de Sept. de 2025 a las 2:08
Comentada: Umar el 14 de Sept. de 2025 a las 3:04
Hello,
I'm currently working with the rlAgentDDPGOptions and while I have a solid understanding of most of the configuration parameters, I am having some trouble understanding three specific options: NumEpoch, MaxMiniBatchPerEpoch, and LearningFrequency.
My main confusion revolves around how and when the networks are updated, considering the agent's sampling times. Specifically:
  • NumEpoch: How does this parameter relate to the overall training process, and how does it affect network updates?
  • MaxMiniBatchPerEpoch: How does this limit the number of mini-batches processed during an epoch, and how does it interact with the sampling process?
  • LearningFrequency: How does this parameter influence the frequency of updates relative to the agent’s sampling rate?
Any clarification on these points would be greatly appreciated!
Thank you in advance for your help!
Best regards,
Fabián.
  5 comentarios
Fabián
Fabián el 14 de Sept. de 2025 a las 1:36
Hi @Umar,
Thank you so much for your response. With your clarifications, I fully understand the implementation of the DDPG algorithm in MATLAB. You deserve a cold beer.
Best regards!
Fabián.
Umar
Umar el 14 de Sept. de 2025 a las 3:04

Hi @Fabian, Haha, I’ll happily take that cold beer — thanks! 😄 Glad to hear the explanation helped clarify things. Feel free to reach out anytime if more questions come up.

Iniciar sesión para comentar.

Respuestas (0)

Categorías

Más información sobre Sequence and Numeric Feature Data Workflows en Help Center y File Exchange.

Productos


Versión

R2024b

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by