代写Introduction to Multi-armed Bandit Algorithms Homework Assignment 1代写留学生Python程序

Introduction to Multi–armed Bandit Algorithms

Homework Assignment 1

Instructions

There are three problems in this Assignment that are related to each other.

• Attach your code to receive full credit. We recommend using Python but the choice is yours.

• It is recommended that you use a Document Scanning application, e.g., iOS Notes, CamScanner, to convert your solutions to PDF format.

Comparing the Performance of Multi-armed Bandit Algorithms

Through a set of assignments, we will learn how to apply multi–armed bandit algorithms in real–life applications.  We will also compare the performance of three multi–armed bandit (MAB) algorithms on real datasets. The datasets we will use to that end are as follows:

• Movie Lens Dataset: https://grouplens.org/datasets/movielens/1m/

• Goodreads Dataset: https://github.com/MengtingWan/chainRec/tree/master

• Open Bandits Dataset: https://arxiv.org/pdf/2008.07146.pdf

Choose one of these datasets.  Movie Lens and Goodreads datasets contain user ratings  for movies and books, respectively.  Your main goal is to compare the performance of the fol–  lowing MAB algorithms:  Explore–then–commit (ETC), Upper Confidence Bound (UCB), and  Thompson Sampling (TS). The comparison will be in terms of the expected cumulative Regret  an algorithm incurs until round t, where t = 1, . . . , n. Here, n defines the horizon, i.e., the total  number of rounds the algorithm is used.  Depending on the data set, decide on what “arms” may represent and what “rewards” may represent.  For example, each movie genre can be an  arm, and user ratings can be considered as the reward received when movie from a genre is  rated by a user. Please stick with the same set–up and data–set for all questions below.

Problem 1

This question aims to demonstrate the fact that theoretical results presented in the lectures are for the mean regret.  Thus, to check their validity in real-world experiments, we need to run a sufficient number of experiments and look at the average value of the cumulative regret.

Consider the ETC algorithm.  Choose the horizon as n = 100, 000 and set the length m * k of the exploration phase as ≈ 10% of n, i.e., m * k ≈ 10, 000. In Problem 3, we will explore the impact of m on the performance. Using your chosen dataset, run the ETC algorithm and record the cumulative regret at each round t = 1, . . . , n. Repeat this process 10 times, reshuffling the data each time before the experiment is run.

Plot the result of ten experiments separately on the same figure. Namely, your figure should have ten curves, one for each experiment, representing the cumulative regret in round t as t varies from 1 to 100, 000. Comment on what you see from these figures. For example, do you get the same regret value in each experiment?

Now, run 100 experiments with the same setting and plot the average regret together with error bars indicating one standard deviation above and below the mean.  See Figure 1 below for an example plot.

For all remaining parts, plots should be obtained by averaging over 100 experiments and errors bar should be shown as indicated above (unless otherwise indicated).

Problem 2

This question is about setting the value of horizon n appropriately.  Some of the theoretical results presented in lectures, e.g., the fact that cumulative regret scales as the logarithm of n for the algorithms presented, are valid in the limit of n → ∞ . In practice, how large n should be to see this logarithmic behavior. may depend on the particular setting used including the algorithm and the reward distributions.

We will consider five different values for the horizon:  n = 500, n = 5, 000, n = 50, 000, n = 500, 000, and n = 5, 000, 000.  Consider the ETC algorithm with the length m * k of the exploration phase set as 10% of n. For each nvalue, plot the average regret of the ETC algorithm as a function of t = 1, . . . , n, i.e., you will produce five figures, one for each n value.  For this question, error bars are not needed.

Comment on the results.

Problem 3

In this question, we will explore how to set the parameter m for the ETC algorithm, i.e., to see how long should we explore.

Set n = 100, 000.  For your chosen dataset, plot the performance of the ETC algorithm for four different m values: m * k = 100, m * k = 1, 000, m * k = 10, 000, m * k = 20, 000. As before, average over 100 experiments and show the error bars. Comment on the results you are seeing.

Figure 1: Example plots showing cumulative regret as a function of the number of rounds for different algorithms.



热门主题

课程名

mktg2509 csci 2600 38170 lng302 csse3010 phas3226 77938 arch1162 engn4536/engn6536 acx5903 comp151101 phl245 cse12 comp9312 stat3016/6016 phas0038 comp2140 6qqmb312 xjco3011 rest0005 ematm0051 5qqmn219 lubs5062m eee8155 cege0100 eap033 artd1109 mat246 etc3430 ecmm462 mis102 inft6800 ddes9903 comp6521 comp9517 comp3331/9331 comp4337 comp6008 comp9414 bu.231.790.81 man00150m csb352h math1041 eengm4100 isys1002 08 6057cem mktg3504 mthm036 mtrx1701 mth3241 eeee3086 cmp-7038b cmp-7000a ints4010 econ2151 infs5710 fins5516 fin3309 fins5510 gsoe9340 math2007 math2036 soee5010 mark3088 infs3605 elec9714 comp2271 ma214 comp2211 infs3604 600426 sit254 acct3091 bbt405 msin0116 com107/com113 mark5826 sit120 comp9021 eco2101 eeen40700 cs253 ece3114 ecmm447 chns3000 math377 itd102 comp9444 comp(2041|9044) econ0060 econ7230 mgt001371 ecs-323 cs6250 mgdi60012 mdia2012 comm221001 comm5000 ma1008 engl642 econ241 com333 math367 mis201 nbs-7041x meek16104 econ2003 comm1190 mbas902 comp-1027 dpst1091 comp7315 eppd1033 m06 ee3025 msci231 bb113/bbs1063 fc709 comp3425 comp9417 econ42915 cb9101 math1102e chme0017 fc307 mkt60104 5522usst litr1-uc6201.200 ee1102 cosc2803 math39512 omp9727 int2067/int5051 bsb151 mgt253 fc021 babs2202 mis2002s phya21 18-213 cege0012 mdia1002 math38032 mech5125 07 cisc102 mgx3110 cs240 11175 fin3020s eco3420 ictten622 comp9727 cpt111 de114102d mgm320h5s bafi1019 math21112 efim20036 mn-3503 fins5568 110.807 bcpm000028 info6030 bma0092 bcpm0054 math20212 ce335 cs365 cenv6141 ftec5580 math2010 ec3450 comm1170 ecmt1010 csci-ua.0480-003 econ12-200 ib3960 ectb60h3f cs247—assignment tk3163 ics3u ib3j80 comp20008 comp9334 eppd1063 acct2343 cct109 isys1055/3412 math350-real math2014 eec180 stat141b econ2101 msinm014/msing014/msing014b fit2004 comp643 bu1002 cm2030
联系我们
EMail: 99515681@qq.com
QQ: 99515681
留学生作业帮-留学生的知心伴侣!
工作时间:08:00-21:00
python代写
微信客服:codinghelp
站长地图