Victor Baybekov

* [ ] Simple moving regression is done as a proof of concept. Not profiled for allocations, performance, etc., just the betas look like correct, but not compared to exact...

The state is not cached, we recalculate lagged xpxi on every move since it is done via zipn, but that is straightforward to fix via manual impl of BindCursor. The...

But we do just twice the work of M size, not N, so this is already online also.

Review https://github.com/dotnet/corefx/pull/31779 Small matrices are the most common case

https://en.wikipedia.org/wiki/Sherman%E2%80%93Morrison_formula https://en.wikipedia.org/wiki/Recursive_least_squares_filter > Compared to most of its competitors, the RLS exhibits extremely fast convergence. However, this benefit comes at the cost of high computational complexity. https://en.wikipedia.org/wiki/Woodbury_matrix_identity > This is...

Why not just `series.Lag(n)`? What are you trying to achieve? A sum of all values or a running/moving sum? You are calculating `series.Lag(1).Lag(1).Lag(1).Lag(1)...` n times. Lag is very optimized for...

Even chained `series.Lag(1).Lag(1).Lag(1).Lag(1)...` should have `O(1)` `MoveNext()` and the goal of the lib is to optimize for it (this is the real-time case when new data points arrive). But `lag(100)`...

Actually lagged cursor is quite complex now and keeps track of current and lagged position, instead of just an offset by N, so it require two cursor moves. Nesting such...

Do you have absolute numbers for your benchmarks?

I really do not understand from your code what are you trying to achieve? How comes that Scan replaces Lag? Also, since you are using the built-in F# Option type...