Description
Question 1
In this question an agent is trained to navogate a grid world, reach the goal state and avoid certain states.
The grid is represented as a nested list where +100 represent a green state and -100 represent a red state. Every other state is marked -1 as a cost of moving 1 step.
the agent either makes a random move with some prodabilty exp which is the exploration part of make the best move, i.e. state with maximum reward with probability 1 −exp.
Then the agent is trained and the expected reward for each state id learned. The final out put is shown with. a nexted list with alphabets ”U”, ”D”, ”L”,
”R”, ”X”, ”W”, where
”U” means the best move from that state is to go up
”D” means the best move from that state is to go down
”L” means the best move from that state is to go left
”R” means the best move from that state is to go right
”X” means that its a red cell
”W” means that its a green cell
Question 2
In this question we used self organizing maos to cluster a data set of world population density.
The output layer is of size 10 × 10, and each input vector is of size 6. The output shows the output layer and each neuron with the weights attachd to that neuron after the clustering is complete.
1
