Meeting chat log
00:10:27 pavitra: hahaha..look at Jon resampling his wine
00:10:38 Jonathan Trattner: HAH
00:10:44 pavitra: 1-fold CV
00:11:16 pavitra: that's bad..jon's test and train sets are the same
00:11:36 pavitra: HAHAHA...
00:12:05 pavitra: Let me grab my IPA. Tan, get your Milk stout
00:12:30 Tan Ho: bahaha bottoms up!
00:13:23 Tan Ho: *awkward eyes meme*
00:13:30 pavitra: lmao
00:30:59 Tyler Grant Smith: i feel like this deserves experimental support to compare 50 fold cv vs 5fold x10
00:34:35 pavitra: gotcha…
00:42:34 pavitra: is this bootstrapping with replacement?
00:43:42 Jon Harmon (jonthegeek): 63.2% 36.8%
00:43:54 pavitra: 2 SDs
00:44:34 Conor Tompkins: @max why is it those numbers
00:45:24 Tan Ho: (or just doubleclick the function lol)
00:45:42 Jonathan Trattner: Both of those are GAME changers
00:45:43 Tyler Grant Smith: ive lost the ink on my f2 key i use it so much
00:45:46 Jonathan Trattner: wowowow
00:45:49 Tan Ho: er, wait, ctrl-click?
00:45:51 pavitra: yeah - I just discovered "go to function definition" in the code menu myself
00:46:31 Jonathan Trattner: Productivity boost 🤯
00:46:42 Tan Ho: it's a trap! - admiral ackbar
00:46:52 Tan Ho: (you can end up digging too deep)
00:47:16 pavitra: tyler is an expert at time series
00:54:03 Tyler Grant Smith: this seems to be contrary to how hyndman defines it
00:58:01 pavitra: isn't there a way to dynamically tune your hyperparameters such that it can do it for the defined range and you can see the performance using the control?
00:59:14 Jim Gruman: time series resampling at the tidymodels web site covers `sliding_window()` `sliding_index()` and `sliding_period()` functions. Maybe the book 10.2.4 rolling forecasting could be a little more clear about all three methods
00:59:31 Tyler Grant Smith: reading the documentation for rolling_origin i think that graphic represents cumulative=FALSE not TRUE. This agrees with the books code
01:06:24 Jon Harmon (jonthegeek): lm_res$.extracts[[1]][[1]]
01:07:34 Tyler Grant Smith: may not be tidy but thats code im used to!
01:09:44 Jonathan Trattner: Go, Jon, go!
01:10:44 Jim Gruman: thank you Jon!!!!!
01:10:52 Tyler Grant Smith: step_interact( ~ Gr_Liv_Area:starts_with("Bldg_Type_")) neat i didnt know you could do that
01:12:29 Tyler Grant Smith: my models are all organic non-gmo
01:17:36 Joe Sydlowski: For Conor's question: http://uc-r.github.io/mars
01:17:47 Jonathan Trattner: Thanks, Jon! Have a great night, everyone!
01:18:14 Joe Sydlowski: There's a plot with model selection for the autoplot in MARs that shows the metrics
01:18:19 pavitra: thanks Jon
01:18:19 Conor Tompkins: Thanks Jon!