Artificial Intelligence/My research results

From Wikibooks, open books for an open world
Jump to: navigation, search

In my quest to construct a circit that mimics a neuron in the brain, I ended with a circuit that exhibitted the following results

There are two types of learning

  • Memorized learning
  • Algorithimization

Memorized learning[edit]

Lets take the case of Fibonacci series 0,1,1,2,3,5,8,13,21,34 .............. Say you want to train a circuit the when given a number n gives out the nth Fibonacci number. Example if given n=5 the circuit gives out 3 and so on. When we try to train the circuit by this way the circuit simply remembers that at n=0,1,2,3,4,5 and so on the output must be 0,1,1,2,3 etc.

Algorithmization[edit]

But instead if we train give a seed value of 0,1 and variable n which represents a feed back we can get a circuit that forms a algorithm that can generate Fibonacci series, rather than one that just remembers.

If a circuit undergoes algorithmized training it can generate output for previously encountered and non encountered values that are given as input. Else if a circit has learned by memorization it can give right output for previously trained values, but may give out erroneous output for previously nonencountered inputs.