1. 4
2. 5
3. 3
4. 3
5. Didn't do
6. Didn't do
7. 6
8. 7
9. 6
10. 5
Go down the most promising state first.
Best first search with a random heuristic would be equivilant to random.
Constant evaluation function would be equivilant to breath first search, search first discovered.
Our heuristic function is inaccurate.
We can get stuck at a certain amount of steps from goal continuasly if we get stuck in a cyle when transversing children.
Algorithim A is similar to best-first except it handles
loops, it accounts for distance from start at each level
with the formula h(n) + g(n) = f(h). Where h(n) is evaluation function and
g(n) is distance from start.
Dinstance from the start will stop an inaccurate heuristic
from progressing too far down in terms of plys, at each ply
1 is added to the g(n) value, when the total output of the formula
exceeds the previous best we backtrack and go down that one instead.
Necessary for genetic algorithim:
- Mutation rate
- Crossover rate
- Population size
- Population of creatures with varied DNA
It has no idea what the global maximum looks like. It could possibly find it over an infinte period of time but even then it can't be sure.
If variable population size is allowed it's possible that the population will grow too large, exceeding expected size (explode) or they could go entirely extinct.
| i | f(i) |
|---|---|
| 1 | 0 |
| 2 | 0 |
| 3 | 0 |
| 4 | 1 |
i)
if i = (1,2,3) then f(i) = 0
if i = (4) then f(i) = 1
- p(1,2,3) = 0.1748
- p(4) = 0.4753
ii)
if i = (1,2,3) then f(i) = 0
if i = (4) then f(i) = 1
- p(1,2,3) = 0.2436
- p(4) = 0.2692
iii)
Limits question, will revise later
i)
if i = (1,3) then f(i) = 0.2
if i = (2,4) then f(i) = 1
- p(1,3) = 0.08993
- P(2,4) = 0.4910
ii)
Limits question, will revise later
;)