(a) NURBS and subdivision are alternative methods for representing…

Question Answered step-by-step (a) NURBS and subdivision are alternative methods for representing… (a) NURBS and subdivision are alternative methods for representing surfaces in three dimensions. (i) Compare and contrast NURBS and subdivision for representing surfaces in three dimensions. [5 marks] (ii) Suggest why the animation industry favours subdivision and the CAD industry favours NURBS. [2 marks] (b) Chaikin’s corner-cutting curve subdivision method is based on the quadratic uniform B-spline. (i) Describe Chaikin’s method. [3 marks] (ii) Describe the quadratic uniform B-spline for representing curves. [4 marks] (iii) Describe how Chaikin’s method for curves is extended and generalised to produce a subdivision method for surfaces that is able to cope with polygons with any number of sides and to cope with vertices with any number of incident edges. A reinforcement learning problem has states {s1, . . . , sn}, actions {a1, . . . , am}, reward function R(s, a) and next state function S(s, a). (a) Give a general definition of a policy for such a problem. [1 mark] (b) Give a general definition of the discounted cumulative reward and the corresponding optimal policy for such a problem. [5 marks] (c) Give an expression for the optimal policy in terms of R, S and the discounted cumulative reward, and show how this can be modified to produce the Q-learning algorithm. [7 marks] In a simple reinforcement learning problem, states are positions on a grid and actions are up and right. The only way an agent can receive a reward is by moving into one of two special positions, one of which has a reward of 10 and the other of ?100. ?100 s1,1 s1,2 s5,3 10 s4,3 Here, states are labelled by their grid coordinates. A possible sequence of actions (sequence 1) is shown by solid arrows, ending with a reward of 10 being received, and another (sequence 2) by dashed arrows ending with a reward of ?100. (d) Assume that all Q values are initialised at 0. (i) Explain how the Q values are altered if sequence 1 is used twice in succession by the Q-learning algorithm. [4 marks] (ii) Explain what further changes occur to the Q values if sequence 2 is then used once by the Q-learning algorithm. [3 marks] (a) Considerable recent Bioinformatics research has focused on phylogenetics. (i) What is the motivation for this work? [1 mark] (ii) Describe with the aid of examples two different techniques for phylogeny. In each case discuss the issues of complexity and performance. [4 marks each] (b) Considerable recent Bioinformatics research has focused on structure prediction from sequence data. (i) Describe how you would build a hidden Markov model (HMM) to identify membrane segments in aminoacid sequences. [6 marks] (ii) How you would assess the sensitivity and specificity performance of your HMM? Image transcription textExercise 10 Doubling Numbers (MA 2+ 2 + 2 + 2+ 2 credits) The following piece of code iscalled Half: X : = 0; y : 0; while (x &l… Show more… Show moreImage transcription textAppendix – Hoare Logic Rules .Assignment: (Q(e) } x := . (Q(x)) .Precondition Strengtheni… Show more… Show more  Computer Science Engineering & Technology Networking ECON 2040 Share QuestionEmailCopy link Comments (0)