optimizate the machine learning/deep learning model creation

[jero98772]

we know if we have more data ,time and power we can do more accuracy predictions, we can do pre-computations and creating one powerful model, but the behavior of each sensor is different
we think how to make a super model , or multiples models for reduce the time of training and make more accuracy predictions

[jero98772]

the problem is the follow,

we need a way to optimize the process to build "strong models" with "strong algortihms" like lstm, rnn, tcn and more
how can we create , or use something that not train from 0 this networks

i think something like this

Agentic Retrieval-Augmented Generation for Time Series Analysis
https://arxiv.org/pdf/2408.14484

but this is a commun problem asociated with other thinks so there is a more simple solution