Conversion of DropSlots, MutualInformationFeatureSelection, and CountFeatureSelection into estimator and transformers #1683
Conversation
| @@ -179,9 +186,52 @@ public bool IsValid() | |||
| } | |||
| } | |||
|
|
|||
| public sealed class ColumnInfo | |||
There was a problem hiding this comment.
ColumnInfo [](start = 28, length = 10)
You can add same summary as you have in constructor #Closed
| Host.Assert(AreRangesValid(SlotsMin, SlotsMax)); | ||
| } | ||
|
|
||
| // Factory method for SignatureLoadModel |
There was a problem hiding this comment.
don't forget dot in the end of sentence :) #Closed
There was a problem hiding this comment.
| private readonly ColumnType[] _dstTypes; | ||
| private readonly SlotDropper[] _slotDropper; | ||
| // Track if all the slots of the column are to be dropped. | ||
| private readonly bool[] _suppressed; |
There was a problem hiding this comment.
why it's one dimensional array? don't we have array of columns and in each column we have slots? #ByDesign
There was a problem hiding this comment.
The ith element of this array is true if all slots are dropped from the ith column. We are given a set of columns and related ranges of slots to drop for each column. This indicates if for the ith column all the slots will be dropped. So this does not indicate if a give slot is dropped (that's what the ranges are for), but rather whether all slots will be dropped.
In reply to: 236089331 [](ancestors = 236089331)
|
Any thoughts @[email protected] ? In reply to: 440747947 [](ancestors = 440747947,440492847) |
| .Append(new CountFeatureSelector(Env, "bag_of_words", "bag_of_words_count", 10) | ||
| .Append(new MutualInformationFeatureSelector(Env, "bag_of_words", "bag_of_words_mi", labelColumn: "label"))); | ||
| var est = new WordBagEstimator(ML, "text", "bag_of_words") | ||
| .Append(ML.Transforms.FeatureSelection.CountFeatureSelectingEstimator("bag_of_words", "bag_of_words_count", 10) |
There was a problem hiding this comment.
ML.Transforms.FeatureSelection [](start = 24, length = 30)
why you need full name? #ByDesign
There was a problem hiding this comment.
I wanted to test that I can access it from mlcontext
In reply to: 236791021 [](ancestors = 236791021)
ghost
locked as resolved and limited conversation to collaborators
Ongoing work on converting the transformers to estimators (#754). In this PR I convert DropSlots, MutualInformationFeatureSelection, and CountFeatureSelection into estimator and transformers with relative extensions.
In particular: