Hi guys!
For your happyness i write other article about ML.NET concept for you!
Today i write about the Prediction function.
Prediction function
You can see the Prediction function like a tool that take a row and apply to it a Transformer.
Once that we obtain the model we could use it for make some prediction using the api model.Transform. We must remember that the model obtained before is a Transformertrained by the api Fit(), or a model we have load in some place.
Usually, in real life, we don’t have a batch where we could make prediction, but we have scattered examples and we need to make fast prediciton on the fly or immediately.
we can reduce this to the batch prediction:
- Create a data view with exactly one row.
- Call model.Transform(data) to obtain the ‘predicted data view’.
- Get a cursor over the resulting data.
- Advance the cursor one step to get to the first (and only) row.
- Extract the predicted values out of it.
The above algorithm can be implemented using the schema comprehension with two user-defined objects InputExampleand OutputPrediction as follows:
predictive function
But this would be cumbersome, and would incur performance costs. Instead, we have a ‘prediction function’ object that performs the same work, but faster and more convenient, via an extension method MakePredictionFunction:
The same predictionFunc can (and should!) be used multiple times, thus amortizing the initial cost of MakePredictionFunction call.
Prediction function is not re-entrant / thread-safe, so if you want to maker simultaneous prediction with more thread, you need a function prediction for every thread.
References: