Question

please answe using typing

what is  the idea that MDPs and Reinforcement Learning are useful procedures in AI

Real life examples and engage in self-reflection, both common practices by researchers developing new AI techniques.

Select a problem using MDPs and/or Reinforcement Learning that may arise in the real world.    

Answer 1

Reinforcement learning -Reinforcement learning is a real life based technique related to artificial intelligence.Using this we learn what to do or what not to do in a particular situation on repeatedly doing that.Reinforcement learning can have 3 main components-State,Activity and Reward.lets us take example in real world of simple robot that is trying to walk on a plane.

State-state is the current situation in learning process.In case of our example the state is the position of the legs while trying to walk safely without any problem.

Activity-Activity is the decision what can we do on a particular state.In our example the robot will decide how much long step should take so that it will not fell down or what should be speed to walk without imbalance.

Reward-Reward means modification is current state after taking particular action.Reward can be positive as well as negative.If it is positive then it is normal and if it is negative then it is punishment .So that will be applied as feedback to the state before applying further action.

In real world complex problems we need to choose the highest reward based on current information.

It is mainly used in gaming,online advertisement ,robotic applications etc.

MPDs-MPD tools are for creating models for the uncertain searching problems for an object.

For example let us say a robot is walking on maze to reach destination.So it need to search the direction left or right from which it will get maximum reward.This we can get using MPD tools.