Listing of neural architecture search for time-series packages

[microprediction]

In other words ... better packages than this one.

[microprediction]

Microsoft NNI includes some really nifty optimizations and a clean separation of the space definition from search strategy. It leans on the retiarii framework under the "one shot" category.

[microprediction]

Keras tuner and on top of that, AutoKeras which includes pipelines.

[microprediction]

ts.gluon has things like DeepAR search.

[microprediction]

Auto-Pytorch and in particular the paper Efficient Automated Deep Learning for Time Series Forecasting.

[microprediction]

darts is the original differentiable architecture search per the paper.

[microprediction]

DeepHyper has been used for spatial time-series here.

[microprediction]

Paper Universal Differential Equations for Scientific Machine Learning describes the goals of DiffEqFlux on top of Flux.