# Tips for improving performance of NODEs with sparsely sampled time series?

**URL:** https://discourse.julialang.org/t/tips-for-improving-performance-of-nodes-with-sparsely-sampled-time-series/103828
**Category:** Modelling & Simulations
**Tags:** time-series, diffeqflux, training
**Created:** [September 13, 2023, 8:34pm UTC](https://discourse.julialang.org/t/tips-for-improving-performance-of-nodes-with-sparsely-sampled-time-series/103828 "2023-09-13T20:34:04Z")
**Posts on this page:** 2
**Page:** 1

<div class="post-metadata">

### Author: ![jarroyoe](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/jarroyoe/32/42482_2.png) [@jarroyoe](https://discourse.julialang.org/u/jarroyoe)
#### Post date: [September 13, 2023, 8:34pm UTC](https://discourse.julialang.org/t/tips-for-improving-performance-of-nodes-with-sparsely-sampled-time-series/103828/1 "2023-09-13T20:34:04Z")

</div>

I have a synthetic time series coming from a model with a daily time scale. The real dynamics closely follow a sine function, but I’m trying to sample at a monthly time scale (every 30 time steps). A sample time series looks like this:  
 ![sparseData](https://global.discourse-cdn.com/julialang/original/3X/b/a/ba941181c950b53758d47669164f3782ae086221.png)

However, when training a NODE model with L2 regularization, the NODE just learns to comfortably follow the mean of the time series:  
 ![sparseDataNODE](https://global.discourse-cdn.com/julialang/original/3X/a/6/a616c87cc6ee9f4d7517d630b35cfd55fa890e36.png)

What would be a good approach to get dynamics closer to the real dynamics when the sampling is this sparse?

---

<div class="post-metadata">

### Author: ![ChrisRackauckas](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/chrisrackauckas/32/77_2.png) [@ChrisRackauckas](https://discourse.julialang.org/u/ChrisRackauckas)
#### Post date: [September 13, 2023, 8:46pm UTC](https://discourse.julialang.org/t/tips-for-improving-performance-of-nodes-with-sparsely-sampled-time-series/103828/2 "2023-09-13T20:46:41Z")

</div>

> [@jarroyoe](#):
>
> What would be a good approach to get dynamics closer to the real dynamics when the sampling is this sparse?

See the tutorial page titled " [Strategies to Avoid Local Minima](https://docs.sciml.ai/SciMLSensitivity/stable/tutorials/training_tips/local_minima/#Strategies-to-Avoid-Local-Minima)"

[https://docs.sciml.ai/SciMLSensitivity/stable/tutorials/training\_tips/local\_minima/](https://docs.sciml.ai/SciMLSensitivity/stable/tutorials/training_tips/local_minima/)