2

I am new to machine learning, so please excuse my level of details here.

I am looking to see if there is a way to develop and train a machine learning model to look at the RDF graph and start detecting graph patterns and flag a particular node if it detects a graph pattern.

I am not really looking to export the data to a text format and run it through a AutoML or similar tool to train and use the model, but as much as possible stick to the rdf store itself for the pattern matching.

Can someone please share me any pointers that can be useful for my need.

Mike Gardner
  • 6,611
  • 5
  • 24
  • 34
Kiran
  • 2,997
  • 6
  • 31
  • 62
  • 1
    Sounds interesting! Unfortunately, this kind of broad research/idea question is not really appropriate for StackOverflow. We can only answer questions related to specific programming problems or tasks. – cygri May 03 '19 at 17:59

1 Answers1

3

since you want to keep data inside MarkLogic, I suggest you try out the upcoming MarkLogic 10 beta, with in-database machine learning capability. For your question, if you can afford to hand label RDF graphs with "patterns" then you can try classification models, otherwise you can try clustering models. These are just starters, for RDF graphs specifically, there might be much more suitable types of models.

For MarkLogic 10, the majority of your work will be preparing training data, and selecting/designing/building/training of the model.

liuyw1938
  • 31
  • 2
  • Thanks @liuyw1938, is there an online documentation that I can refer to to understand how to use this? – Kiran May 06 '19 at 08:19
  • @Kiran You can start with [cntk](https://cntk.ai/pythondocs/) which is Microsoft's documentation on their machine learning toolkit, CNTK. This is what gets embedded into MarkLogic 10. The usage will be similar in MarkLogic as to in Python. For RDF specific stuff, you have do your own research. – liuyw1938 Jun 01 '19 at 20:15