| Age | Commit message (Collapse) | Author | Lines | 
|---|
|  |  | 
|  |  | 
|  |  | 
|  |  | 
|  |  | 
|  |  | 
|  |  | 
|  |  | 
|  |  | 
|  | get_fitted_sub is no longer present, sub-state models are part of get_fitted now | 
|  |  | 
|  |  | 
|  |  | 
|  |  | 
|  |  | 
|  |  | 
|  |  | 
|  |  | 
|  |  | 
|  | use application-provided total state duration instead of model to determine
duration of last sub-state in get_fitted_sub | 
|  |  | 
|  |  | 
|  |  | 
|  | TODO: fitting and submodel usage in parent | 
|  | Iteration over states/transitions and model attributes is no longer hardcoded
into most model generation code. This should make support for decision trees
and sub-states much easier. | 
|  |  | 
|  | Note that outliers are removed from both training and evaluation data | 
|  |  | 
|  | TODO: Fix send duration model generation for cc1200, should improve eval | 
|  |  | 
|  |  | 
|  |  | 
|  |  | 
|  |  | 
|  |  | 
|  | this also fixes a nasty off-by-one in the barcode loader
(now _load_energytrace):
wrong:   interval_start_timestamp = data[:-1, 0] * 1e-6
correct: interval_start_timestamp = data[1:, 0] * 1e-6 | 
|  |  | 
|  |  | 
|  |  | 
|  |  | 
|  |  | 
|  |  | 
|  |  | 
|  |  | 
|  |  | 
|  |  | 
|  |  | 
|  |  | 
|  |  | 
|  |  |