Is your feature request related to a problem? Please describe.
In general, trainers such as KMeans, expect the feature vector to be a vector of known size. The quickest way to provide this information is in the class that defines the model input schema, you can use the [VectorType(n)] attribute where n is the size of your vector. Alternatively, you can use SchemaDefinition similar to this sample. In this second approach, you still need to have an object to define the schema of objects.
Describe the solution you'd like
For dense vectors where the size is the same, there's two approaches we might consider:
Automatically detect the vector size.
Enable users to specify at runtime that a column / property is a vector without using SchemaDefinition.
Is your feature request related to a problem? Please describe.
In general, trainers such as KMeans, expect the feature vector to be a vector of known size. The quickest way to provide this information is in the class that defines the model input schema, you can use the
[VectorType(n)]
attribute wheren
is the size of your vector. Alternatively, you can useSchemaDefinition
similar to this sample. In this second approach, you still need to have an object to define the schema of objects.Describe the solution you'd like
For dense vectors where the size is the same, there's two approaches we might consider:
SchemaDefinition
.Describe alternatives you've considered