1.What is Learning?
Learning is the process of acquiring new, or modifying existing, knowledge, behaviors, skills, values, or preferences. We develop our high-level skills such as reasoning and problem solving thanks to our experiences. Experience is, the change of the state of an organism to be more successful in the event of subsequent situations.
There are 3 types of learning:
- Habit
- Associative learning (classical conditioning, Operant Conditioning)
- Observational Learning
1.1. Habit
Habit is the simplest type of learning. Learning to ignore a stimulus unless it has serious consequences. When a stimulus does not interest the person, the severity of reaction decrease.
1.2 . Associative learning
1.2.1. Classical Conditioning
Classical conditioning, learning specific events which are occurring together. The reaction of the animals was the subject of the Russian physiologist Pavlov’s experiment. Pavlov predicted the dogs would salivate in response to the food placed in front of them, but he noticed that his dogs would begin to salivate whenever they heard the footsteps of his assistant who was bringing them the food.
1.2.2. Operant Conditioning
Operant Conditioning is learning the behavior that will reward or prevent punishment. The person takes action to achieve the desired result without waiting for the warning ( for more information, you can look psychologist B.F. Skinner’s experiments).
1.3. Observational Learning
Observational learning occurs through observing the behaviors of others and imitating those behaviors — even if there is no reinforcement at the time. Albert Bandura noticed that children often learn through imitating adults, and he tested his theory using his famous Bobo-doll experiment. Through this experiment, Bandura learned that children would attack the Bobo doll after viewing adults hitting the doll.
So how is learning on computers and is it possible to model it?
Learning definition in artificial systems is the system is able to perform a work that is not known before after a certain training. Today, the most researched learning style is learning model according to examples.
Multi feedback-Layer Neural Network
We request the child not to eat the hot pepper on the table. That makes him curious .He takes the hot pepper in his mouth and then he understands what the warning means. This learning method is multi-feed back learning. To sum up,people learn with the help of their experiences. Computers work the same way. Let’s explore how the learning algorithm works on the example, according to in this algorithm, system want to reach some object features.(yes-no questions). System changes the decision-making mechanism when the user does not confirm the decision. The system continuously improves itself during the learning process.
Let's examine how to teach the classification of objects to the computer. Our goal is to estimate which class the object is belongs to. For instance, lets see artificial system how distinguishes plane and bird.
>Bird and plane have wings and tail.
>Bird has beak and feathers.
>Plane has chassis and engine.
The first list will include the features of the bird, the second list the features of plane, and third list have common features of them.
POSSIBLE RESULTS: bird or plane
DOES THE OBJECT CONTAIN THIS PARAMETER?(Y/N)
wings 1 (y), tail 1(y), feathers 0 (n), beak 0 (n), chassis 1(y),engine 1 (y)
The first list will be empty, because of the system does not have any information about the bird. Likewise, the system has no knowledge about plane so it will be empty too. Since both comparison results are zero, the system estimates that the entered values will correspond to list 1. System prompts the user to confirm the prediction. Answer will be “no” ,these fetures belogs to the plane. Feedback mechanism will be activated because the result is incorrect. System will fill the list 1 with zeros,because we know answer is not bird. The values entered will be written to the list 2 as the properties of the plane.
The next step is, When the values added to the first list are the same as the values of the second list, these values will be deleted from the first and second list, added to the common list.
The system asks for confirmation of its prediction until it has completed the learning process.
The creation of a learning system can be summarized as follows:
- The system only changes its values with the feedback mechanism when it encounters opposite information.
- The system compares the entered information with the previously saved information and matches them by searching for the closest similarity.
