Is ARIMA really distributed?

[br...@bigr.io]

I am looking through the source code. I realize that there is this TimeSeriesRDD class which seems to have some data prep functions. However, in the main model classes such as ARIMA, it uses the local Vector class as input, and uses regression methods from the math3 library. Excuses my ignorance, but it appears it is not at all a distributed implementation. 

Please enlighten me how is this a Spark lib? Is it only that limited TimeSeriesRDD part?

[puj...@gmail.com]

I'm afraid the whole lib is not distributed as you can find collect() function in all models provided. 

在 2017年12月19日星期二 UTC+8上午10:38:26，br...@bigr.io写道：