Skip to content

Minibatches #38

@dan-sprague

Description

@dan-sprague

Hey guys,

I'm working with a large dataset with a relatively large number of parameters (last-layer approximation for a neural network). Out-of-the-box VI is simply a non-starter here.

To perform parameter updates in mini-batches, is scaling the contribution of the minibatch to the log-likelihood the primary change?

Thanks for your work!

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions