CA5 Multi-agent Systems

CA5 Multi-agent Systems – Question 5 from the 2017 sample paper

Using question 5 from the sample paper Oct 2017 use the simulated data set from Question 1 on the same paper to:

Part a) Adopt a centralised scheme to all agents and sketch the graphical scheme.

Question 1 described a financial system, with four local agents analysing their own data sets. The graph of the centralised scheme is as follows:

To resolve multi-agent system problems, think of them as different samples, each performing an individual t-test. Then for the multi-agent, consider how you put those samples together to calculate the averages. In question 1 we have four samples.

Part b) Compute the normalised weights and find the global arithmetic mean. Please compute the global solution using R.

We calculate the mean of each of the 4 ‘samples’ or local agents.  Normal distribution was specified with N(j,16) therefore the samples were calculated using rnorm()  function, with a variance of 4.

We now take the samples and calculate the mean for each data set. We create a new sample, which we called xvect below. This is the mean of the 4 samples.

We calculate our weights for each agent:

Below you can see how we can calculate the global arithmetic mean.

We normalise the weights and then use these values to calculate the arithmetic mean. The final calculation for the arithmetic mean is 2.312.