Update trained model and choose own weighting for different columns

[Toudik-Isavi]

Hello,

I have read the documentation and also tried out Zingg but I still have two questions left.

• I wonder if it is possible to update an already trained model with new data?
• Is it possible to choose my own weighting on columns and with that decide how much the columns contribute to the z_score? I would like to train a model with this approach but I do not whether it is actually possible.

It would be awesome if anyone knows the answers to my questions and helps me out.

[sania-16]

• You can add new data using - https://docs.zingg.ai/latest/stepbystep/createtrainingdata/addowntrainingdata or run findTrainingData and label again to update the model with new data - after this you can re-train the model
• No, it is not possible to choose your own weights on fields as Zingg learns the appropriate field weights based on how the user labels the data. So, depending on whichever fields you need to assign more weightage to, the labelling needs to be done for your model. Just to understand better, what is the purpose of this?

[Toudik-Isavi]

Thank you so much for your fast reply, you helped a lot! The purpose of this is simple: I would like to include prior knowledge. This can be especially useful in cases where not so many training samples are available. Of course in other cases with much more training data the prior will be outmatched and it won't play a big role anymore.

[sonalgoyal]

Thanks for the details! Id the trainingSamples functionality outlined in the earlier answer good for you?

[sania-16]

closing as resolved