save_theta produces NULL in the history data result

[yanisvdc]

Hi, thank you for this package,

When using

library(contextual)

horizon       <- 100L
simulations   <- 100L

bandit        <- ContextualLinearBandit$new(k = 4, d = 3, sigma = 0.3)

# Linear CMAB policies comparison

agents <- list(Agent$new(EpsilonGreedyPolicy$new(0.1), bandit, "EGreedy"),
               Agent$new(ContextualEpsilonGreedyPolicy$new(0.1), bandit, "cEGreedy"),
               Agent$new(ContextualLinTSPolicy$new(0.1), bandit, "LinTS"),
               Agent$new(LinUCBDisjointOptimizedPolicy$new(0.6), bandit, "LinUCB"))

simulation     <- Simulator$new(agents, horizon, simulations, do_parallel = TRUE, 
                                save_context=TRUE, save_theta=TRUE)

history        <- simulation$run()

plot(history, type = "cumulative", rate = FALSE, legend_position = "topleft")

result <- history$data

Why are some of the thetas in the result NULL? I believe that there is no theoretical reason for this and that it should be an error when writing the thetas in the result. Would anyone know how to fix this?

Thanks!

[yanisvdc]

Solved this by using do_parallel = False

simulation     <- Simulator$new(agents, horizon, simulations, do_parallel = FALSE,
                                save_context=TRUE, save_theta=TRUE)

It seems to be an issue with how parallelization returns the results.