Learning in Problem Solving with Example in AI

Learning in Problem Solving

Experience can help solve a problem. Experience boosts the efficiency of solving a similar task. Usually, this doesn't need rearranging the data or recalling how to get the answer.

Consequently, learning to solve problems entails: 

  • Acquiring problem-solving skills by personal experience—in the absence of a teacher or advisor. 
  • focuses only on how to use the knowledge; it does not include gaining more knowledge. 
  •  Although learning rules can help focus the issue-solving process, it comes at a cost because the problem solver must remember and refer to those rules frequently.

The software can learn by expanding from its own experiences in situations where it is unable to learn from guidance.

➢ Learning through changing the parameters 

Acquiring knowledge with macro-operators 

Chunking learning

These are three methods by which a system can learn from its own experience.

1. Learning through changing the parameters

An evaluation procedure is used by many systems to summarize the status of the search, among other things. Numerous instances of this can be seen in game-playing apps. 

Still, a static evaluation function is present in many applications.

A small adjustment to the problem evaluation's phrasing is necessary during the learning process. 

In this case, the evaluation function of the problem is represented by a polynomial of the following form:


The c terms are weights, whereas the t terms are feature values. 

Determining the precise value to assign each weight at the outset can be challenging when building programs. 

Thus, the fundamental concept behind parameter adjustment is to: 

Begin by estimating the appropriate weight parameters. Program weight should be adjusted based on.

2. Acquiring knowledge of macro operator

This concept is essentially the same as rote learning. It keeps costly recomputation at rest. You can use macro-operators to combine many actions into a single one.

In this knowledge: 

  • Rote learning as a series of steps has been found to help address problems. 
  • Macro operations are a series of tasks that can be handled collectively. 
  • The learning component takes the computed plan and stores it as a macro-operator after an issue is solved. The initial conditions of the problem that was just solved are known as the preconditions, and the goal that was just accomplished is known as the postconditions.
  • The issue solver makes effective use of the foundation of information it has built from its prior experiences. 
  • The issue solver can even handle different difficulties by expanding the scope of macro-operators.

3. Chunking learning

Chunking is the practice of performing a certain activity by combining a series of actions and treating them as a whole.

Chunking and learning with macro-operators are comparable. It is typically employed by systems for solving problems that require production systems.

An IF-THEN set of rules is what makes up a production system. Production rules are used by the SOAR system to represent its knowledge. Chunking is another technique it uses to gain experience. These works of art are kept in the long-term memory. SOAR generates chunks when it finds a good firing sequence. Before storage, chunks may be generalized.

Post a Comment

* Please Don't Spam Here. All the Comments are Reviewed by Admin.

Top Post Ad

Below Post Ad

Ads Section