summaryrefslogtreecommitdiff
path: root/lib/model.py
AgeCommit message (Collapse)AuthorLines
2021-08-31model: allow leaf function setting to be overridden via envDaniel Friesel-1/+4
Here, it defaults to enabled and can be overridden to disabled (DFATOOL_DTREE_FUNCTION_LEAVES=0)
2021-08-31model: use mean stddev for fixed parameter configuration as split threshold ↵Daniel Friesel-1/+6
by default
2021-08-24support override functions in dtree codeDaniel Friesel-1/+3
2021-08-23move build_dtree method back to ModelAttribute classDaniel Friesel-114/+1
2021-08-23remove unused legacy attr.build_dtree codeDaniel Friesel-5/+0
model.build_dtree does belong into attr, but that's a different commit
2021-08-23Automatically build simple analytic or tree-based analytic model as neededDaniel Friesel-8/+21
The tree variant is used for attributes which depend on a non-numeric parameter (which can't be modeled as a function)
2021-08-20model/dtree generation: detect int parameters, do not hardcode "batch_size"Daniel Friesel-6/+8
2021-08-20model: support for decision tree with function leavesDaniel Friesel-6/+41
pretty hacky at the moment, but good enough for eval
2021-08-19ParallelParamFit -> ParamFit, optionally without parallelismDaniel Friesel-3/+3
2021-07-28model: improve error messageDaniel Friesel-1/+5
2021-06-25model: remove debug statementDaniel Friesel-3/+0
2021-06-23simplify dtree builderDaniel Friesel-25/+52
2021-06-23move kconfig-dtree builder to AnalyticModel. WIP.Daniel Friesel-1/+78
2021-05-10analyze-config, to_json: use paramNames in exported json nodesDaniel Friesel-4/+4
2021-04-20proper kconfig-webconf exportDaniel Friesel-0/+7
2021-04-20add dot export of energy modelsDaniel Friesel-0/+13
2021-03-25model: allow stats generation to be skippedDaniel Friesel-2/+6
2021-03-24model.to_json: Estimate transition origin/destination from traces if unknownDaniel Friesel-4/+33
2021-03-24Add nearly correct JSON export for v0 measurements without PTA definitionDaniel Friesel-0/+5
Transition source/destination names are not correct yet
2021-03-17always handle co-dependent parametersDaniel Friesel-2/+2
2021-03-16move codependent parameter detection to Model / ModelAttributeDaniel Friesel-1/+4
Still TODO: Ignore codependent parameters when partitioning data for analytic modeling / regression
2021-03-08model.assess: remove superfluous "by_name" indirectionDaniel Friesel-11/+8
2021-03-08add dataref exportDaniel Friesel-0/+42
2021-03-05promote sub-state models to a first-class modeling citizenDaniel Friesel-64/+23
get_fitted_sub is no longer present, sub-state models are part of get_fitted now
2021-03-04Restore --export-energymodelDaniel Friesel-2/+1
2021-03-03Remove ModelInfo; add info to ModelFunction insteadDaniel Friesel-4/+4
2021-03-02ModelAttribute + sub-classes: add to_json methodDaniel Friesel-3/+14
2021-03-02PTAModel: Turn states, transitions, and states_and_transitions into attributesDaniel Friesel-33/+24
2021-03-02AnalyticModel, PTAModel: refactor assess()Daniel Friesel-21/+9
2021-03-02ModelAttribute: remove get_fitted(), use .model_function, .model_info insteadDaniel Friesel-6/+7
2021-03-02Move ParallelParamFit to a separate moduleDaniel Friesel-322/+3
2021-03-02move ModelAttribute to parameters moduleDaniel Friesel-247/+6
2021-03-01Set DFATOOL_NO_PARAM to disable parameter fittingDaniel Friesel-0/+2
2021-03-01get_fitted: also provide information on static (sub)modelsDaniel Friesel-2/+5
2021-02-26refactor param_info, show splits in analyze-archive outputDaniel Friesel-28/+17
2021-02-26add simple decisiontree testDaniel Friesel-1/+1
2021-02-25kinda proper dtree support (todo: refactoring)Daniel Friesel-7/+30
2021-02-25pelt_refine: Fix "no sub-states" handlingDaniel Friesel-1/+1
2021-02-25Adjust ParamStats interface in preparation for decision-tree modelsDaniel Friesel-23/+166
2021-02-24cluster and re-fit sub-statesDaniel Friesel-72/+91
2021-02-24lut_median_getter: handle param tuples, not just listsDaniel Friesel-1/+4
2021-02-24do not hard-code substate namesDaniel Friesel-8/+11
2021-02-24treat last sub state as tail / default stateDaniel Friesel-7/+16
use application-provided total state duration instead of model to determine duration of last sub-state in get_fitted_sub
2021-02-23draft for clustering of parameter-independent sub-statesDaniel Friesel-91/+96
2021-02-23analyze-archive: show substate model quality in tableDaniel Friesel-1/+15
2021-02-22add simple sub-state model accessor and evaluationDaniel Friesel-2/+34
2021-02-22properly handle static submodel attributesDaniel Friesel-27/+36
TODO: fitting and submodel usage in parent
2021-02-22PELT: Increase parallelismDaniel Friesel-9/+13
2021-02-19start modeling of individual sub-statesDaniel Friesel-16/+93
2021-02-18refactor model generation from Analytic/PTAModel into ModelAttribute classDaniel Friesel-309/+210
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.