0

Fuzzy Logic allows the user to define rules and determine the output based on the rules and membership functions. It does not require a data-set for learning (I know it doesn't learn. There is a learning version called neuro-fuzzy-systems). I find it very useful in some of my projects because I have more control when using fuzzy logic than using a machine learning algorithm. For example when I want something to work based on certain rules, but still with some uncertainty and intelligence.

What I want to know is, are there any similar AI/Machine Learning techniques in which I can define some rules (as guidance) but also it should have some intelligence beyond some if-else rules. Is there any other techniques similar to fuzzy logic?

Deepan
  • 95
  • 7

1 Answers1

2

Yes there is.

You should have a look at for example https://dtai.cs.kuleuven.be/problog/ Problog combines the programming language prolog with probabilities. this allows you to formulate thinks like

0.3 :: cancer :- smokes. 
0.2 :: cancer :- drinks.
0.1 :: cancer.

You can chain the rule

0.2 :: dead :- cancer

and you could even use variables, however, semantics gets a bit more tricky.

The neat thing is that problog also allows form learning the probability.

You could also use alchemy from the university of washington (Pedro Domingos). It is based on First order logic(1). However, it uses weights instead of probability which is hard to interpret.

(1) the author claim it is full FOL, however, this is not true due to the missing/wrong treatment of functors.

CAFEBABE
  • 3,983
  • 1
  • 19
  • 38