Published May 7, 2025 | Version v6.1.0
Software Open

FlorianKrach/PD-NJODE: Release for Direct Variance Prediction with the Neural Jump ODE model

Creators

  • 1. UC Berkeley

Description

Updated the loss functions, training framework and model to allow for directly learning to predict the variance with a variance loss, using an approach that hard-codes the variance positive-semi definiteness (non-negativity) constraints.

Files

FlorianKrach/PD-NJODE-v6.1.0.zip

Files (177.5 kB)

Name Size Download all
md5:5fc0b92a90dcbf6d27795928b9bd7e88
177.5 kB Preview Download

Additional details

Related works

Is supplement to
Software: https://github.com/FlorianKrach/PD-NJODE/tree/v6.1.0 (URL)

Software

Repository URL
https://github.com/FlorianKrach/PD-NJODE