PcaTrainer as estimator #996
Conversation
| @@ -108,6 +137,18 @@ public override PcaPredictor Train(TrainContext context) | |||
| } | |||
| } | |||
|
|
|||
| private static SchemaShape.Column MakeWeightColumn(string weightColumn) | |||
There was a problem hiding this comment.
Use trainerUtils #Resolved
src/Microsoft.ML.PCA/PcaTrainer.cs
Outdated
| } | ||
|
|
||
| protected override BinaryPredictionTransformer<PcaPredictor> MakeTransformer(PcaPredictor model, ISchema trainSchema) | ||
| => new BinaryPredictionTransformer<PcaPredictor>(Host, model, trainSchema, _featureColumn); |
There was a problem hiding this comment.
a, _featu [](start = 84, length = 9)
@zruty do we have a threeshold, to pass? #Resolved
There was a problem hiding this comment.
| public IDataView Transform(IDataView input) | ||
| { | ||
| Host.CheckValue(input, nameof(input)); | ||
| return Scorer.ApplyToData(Host, input); |
There was a problem hiding this comment.
Scorer [](start = 19, length = 6)
SDCA, FastTree, LightGbm, GAM don't have Scorer setup, so tests are failing. #Resolved
src/Microsoft.ML.PCA/PcaTrainer.cs
Outdated
| @@ -41,7 +42,7 @@ namespace Microsoft.ML.Runtime.PCA | |||
| /// <remarks> | |||
| /// This PCA can be made into Kernel PCA by using Random Fourier Features transform | |||
| /// </remarks> | |||
| public sealed class RandomizedPcaTrainer : TrainerBase<PcaPredictor> | |||
| public sealed class RandomizedPcaTrainer : TrainerEstimatorBase<BinaryPredictionTransformer<PcaPredictor>, PcaPredictor> | |||
There was a problem hiding this comment.
BinaryPredictionTransformer [](start = 68, length = 27)
So, why is this a binary prediction transformer? Prior to your change it was not a binary predictor, it was part of the anomaly detection task. #Closed
There was a problem hiding this comment.
| { | ||
| return scoreType == NumberType.Float; | ||
| } | ||
| => scoreType == NumberType.Float; | ||
| } |
There was a problem hiding this comment.
Bit of a nit, but while you're at it, maybe indent properly. #Closed
|
Closing and re-opening to trigger rebuild. The local build is fine.. |
| /// <param name="rank">The number of components in the PCA.</param> | ||
| /// <param name="oversampling">Oversampling parameter for randomized PCA training.</param> | ||
| /// <param name="center">If enabled, data is centered to be zero mean.</param> | ||
| /// <param name="seed">The seed for random number generation.</param> |
There was a problem hiding this comment.
seed [](start = 25, length = 4)
why do we use a separate seed here? #Pending
| var pipeline = new RandomizedPcaTrainer(Env, featureColumn, rank:10); | ||
|
|
||
| TestEstimatorCore(pipeline, data); | ||
| } |
There was a problem hiding this comment.
} [](start = 7, length = 2)
Done() #Resolved
ghost
locked as resolved and limited conversation to collaborators
Addresses part of #754