Beim Durchlaufen dieses Codes:

simulated HMM -----------------------------------------------------------

seed = 1
controls = list(
states = 3,
sdds = "gamma",
horizon = 500,
fit = list("runs" = 100)
)
controls %<>% set_controls
data = prepare_data(controls, seed = seed)
data %>% summary
data %>% plot
model = fit_model(data, ncluster = 1, seed = seed) %>%
decode_states %>%
compute_residuals
summary(model)
model %<>% reorder_states(state_order = 1:3)
compare(model)
model %>% plot("ll")
model %>% plot("sdds")

wird der 3. Status leider nicht richtig erkannt. Ich habe dasselbe auch mit 1000 Runs einmal ausgeführt, geändert hat sich am Ergebnis allerdings nichts.

The function set_f should

• check inputs
• return function with set arguments

Implement function fixed_initialization:

• run optimisation with fixed input starting values (which can also be a vector of multiple fixed starting values)
• check if the number of inputs matches the number of parameters of the function

Implement a function that captures and returns trace of numerical optimization. For example, the progressive function values, parameter values, steps, gradients, Hessians from nlm().

The function set_optimizer should

• check inputs
• return optimizer with fixed arguments

Replace any set.seed() calls by a call to ino_seed() with a status message if verbose = TRUE.

Implement unit tests for subset_argument(), currently missing.

• add column to summary() output for global optima to simplify the filtering
• add figure with Viterbi decoding of time series that shows why subset_last performs worse
• check references

For the em optimizer, one run converged to value = 248. Explain.

• count number of optima
• add visualisation

After $optimize() and then $clear("all"), $print() throws error. Check!

Random initialization of f with optimizer based on inputs runs, lower and upper (as vectors?) and sampler ("normal", "uniform")

Function that measures the optimization time of different methods and returns overview.

Maybe density plots? With brackets to compare medians? And remove log-scale, this is misleading.