0

For my masters, I'm comparing different approaches to solve a problem (like image segmentation).

I'm comparing machine learning and deep learning approaches to «classical» algorithms (like simple or adaptative threshold).

How would you call these «classical» or «standard » approaches? Like algorithms that don't rely on data but are «hand-designed». I can't seem to find a correct term.

Mohamed Bahdine
  • 1
  • 1
  • 1
  • 1
    maybe rule-based models or rule engine systems? Possible duplicate of https://datascience.stackexchange.com/questions/22545/rule-engine-vs-machine-learning and also see https://www.linkedin.com/pulse/data-science-machine-learning-vs-rules-based-karthik-guruswamy/ – TwinPenguins Jul 31 '18 at 20:40

2 Answers2

1

Machine learning models work on principle of probabilistic approach where you try to fit the function to map input with output. There are no hard bound rules that classify the output based on input.

Deterministic approach like if else are categorized as non machine learning algorithms where every time it follows the same rules to evaluate the results.

CodeMaster GoGo
  • 778
  • 1
  • 6
  • 15
0

We usually refer to non-ML systems as 'expert rule systems'. This is because you are relying on the judgement of an expert, not a system that learns. Ex: you can have a decision tree where thresholds are learned by a machine, or where they are manually assigned using a person's expert knowledge in the criteria.

Chris.Caldwell
  • 101