Computer Science And Information Technology Research (Other (Not Listed) Sample)

Machine Learning
By Name
Institution of Affiliation
Course
Date
Machine Learning involves an agent taking actions within an environment that would enable maximization for a reward in the long term and find the policy that is capable of mapping states to the actions that the agent will take in those states.
Question 1: Definitions
State: It refers to the set of agents and environments that give a vivid illustration of where the action mapping will take place.
Action: It refers to the characteristic way of identifying the basic problems of a computer system to initiate computational cognitive science and artificial intelligence. It is associated to the animats and intelligent agents.
Reward: It refers to the interpretation of the action in a computer machine learning environment to give a representation of the entire state that is taken back to the agent of the machine learning. It is the scalar immediate outcome of the transition process in the machine learning for basic reinforcement.
Question 2:
A). Using Bellman Equation To Obtain The Values Of V’ To 2 Decimal Places.
The state at the top left = 1
V(s)=r(s)+β∑s′p(s′∣s)r(s′)
Maximization state is given by ∑tβtu(ct)
The constraints of the state is given by at+1=1β(at+yt−ct)
Converting to a stochastic equation, we get 1β=Et[u′(ct+1)u′(ct)] which is formulated by the λt factor To give yt+1=yt+ϵt
Hence, the value of V’ is yt+1=yt+ϵt = 10*2 + 1*2 = 22
Top right = 2. the value of V’ is yt+1=yt+ϵt = 0*2 + 0*2 = 0
Bottom left = 3. the value of V’ is yt+1=yt+ϵt = 0*2 + 5*2 = 10
Bottom right = 4. the value of V’ is yt+1=yt+ϵt = 1*2 + 1*2 = 4
B). Using policy iteration to identify the optimal policy
Step 1: Identify the optimal policy using policy iteration at a discounted factor of γ=0.1 and continue with a larger number from 0.99 to 0.9 and then using a smaller number such as 0.01. start with a random policy  INCLUDEPICTURE "/afs/cs/project/jair/pub/volume4/kaelbling96a-html/img55.gif" \* MERGEFORMATINET 

for each iteration while evaluating each policy  INCLUDEPICTURE "/afs/cs/project/jair/pub/vo...