Lecture NINE.

What are the advantages of local search algorithms?

• They always find the global maximum

• They use very little memory

• They are suitable for small state spaces

• They keep track of all the paths

• They are suitable for large or infinity state spaces

Which search algorithm is called greedy local search.

• Greedy First search

• Depth-first search

• Hill Climbing Search

• A* algorithm

What is the main drawback of hill climbing search?

• continually moves in the direction of increasing value

• It requires a little of memory

• It can get stuck in local maxima

• It always finds the global maximum

• Starting from x = 0 , The answer is f(-1) = 3
  ---------------------
  Starting from x = -4 , The answer is f(-2) = 3

• Starting from x = 0 , The answer is f(-2) = 6
  ---------------------
  Starting from x = -4 , The answer is f(-1) = 6

• Starting from x = 0 , The answer is f(-1) = 0
  -------------------
  Starting from x = -4 , The answer is f(-2) = 0

• Starting from x = 0 , The answer is f(0) = -1
  -------------------
  Starting from x = -4 , The answer is f(-1) = -2

What is the main solution to the problem of Local maxima in hill climbing search?

• A big jump

• Random restart

• Allowing 'bad' moves

• Decreasing the temperature

What is the main solution to the problem of plateaus in hill climbing search?

• A big jump

• Random restart

• Allowing 'bad' moves

• Decreasing the temperature

What is the advantage of Simulated annealing search over Hill climbing search?

• allowing some intelligent moves to escape the local maxima

• It always finds the global maximum

• It never gets stuck in local maxima

• To escape local maxima by accepting worse solutions.

What is the main idea behind simulated annealing search?

• Always accept states with better objective function

• Never accept states with worse objective function

• Always take uphill steps

• Gradually decreasing over time as the temperature decreases

if If ΔE > 𝟎; The probability of accepting a state with a better objective function will be :

• T/ΔE

• e^(ΔE)

• 1

• e^(ΔE/T)

if If ΔE < 𝟎; The probability of accepting a state with a better objective function will be :

• T/ΔE

• e^(ΔE)

• 1

• e^(ΔE/T)