The following are known issues in the Splunk Machine Learning Toolkit version 4.2.0.
|Date filed||Issue number||Description|
|2019-05-14||MLA-3100||"Publish" link on the Experiments list page is enabled for Experiments that have never fit a model|
|2019-05-13||MLA-3070||KernelPCA/KernelRidge only gets up to 1000 results|
|2019-05-08||MLA-3064||If _time exists as a column, it shows up as a column in the score's output.|
|2019-03-28||MLA-2965||"Edit Job Settings" does not work in Assistants|
|2019-03-24||MLA-2945||DensityFunction does not support the "as" clause|
|2019-03-19||MLA-2934||Need to document the semantics of DensityFunction parameters in mlspl.conf.spec|
|2019-03-13||MLA-2920||DensityFunction - IsOutlier and BoundaryRanges are blank if split by fields are numeric|
As long as one of the split by fields are non-numeric all should work fine.
| inputlookup my-data.csv | eval field2="_".field2 | fit DensityFunction field1 by "field2,field3,field4" into MyDensityFunction
|2019-01-22||MLA-2818||Experiment Alerts created before MTLK 4.2.0 may be inaccurate on re-save if the user has modified the trigger condition|
If the user has changed the trigger condition of their Experiment alerts from the one originally set when the alert was created, they should visit the Experiment's alerts (via Manage -> Manage Alerts) on the Experiments list after saving the experiment for the first time in MLTK 4.2.0 and confirm that the alert is configured as expected.
|2018-09-24||MLA-2547||Experiments can't be renamed until they have been filled and saved|
Load the Experiment, fill all necessary fields, and save it. You'll be able to rename the Experiment in the Save dialog.
|2018-06-12||MLA-2258||TFIDF returns a vague error message when the "token_pattern" parameter is invalid|
|2018-04-18||MLA-2147||search.log does not show tracebacks on Windows on Splunk 7.1|
The user can still look at the logs on windows in mlspl.log which captures the traceback. However, many users may not have access to mlspl.log (as it is indexed in the _internal index which normal users don't have by default). If this is the case, they will need to ask their Splunk administrators for access to this index.
|2017-08-25||MLA-1916||Uncaught StatsModels error with ARIMA order=0-2-0|
|2017-02-28||MLA-1670||"Outliers Over Time" plot doesn't display any data if the _time field does not contain a Unix timestamp.|
|2016-12-13||MLA-1485||In the Forecast Time Series assistant, forecasting a very large number of points may cause the browser to crash.|
|2016-08-24||MLA-1145||Machine Learning Toolkit does not load correctly if its folder is renamed.|
This documentation applies to the following versions of Splunk® Machine Learning Toolkit: 4.2.0