initial_weightsargument is replaces by
inittakes a named list and currently
R0the initial weights and the initial cumulative regret can be provided. They have to be PxK or 1xK.
profocfunction was extended:
regretcan now be passed as an array as before, or as a list, e.g.
list(regret = regret_array, share = 0.2)if the provided regret should be mixed with the regret calculated by online.
losscan also be provided as a list, see above.
batchfunction can now minimize an alternative objection function using the quantile weighted CRPS
profocfunction was renamed to
batchfunction to apply batch-learning.
oraclefunction to approximate the oracle.
The spline functions where rewritten to add the ability of using a non-equidistant knot sequence and a penalty term defined on the Sobolev space. This change induces breaking changes to small parts of the API.
ndiffdefines the degree of differencing for creating the penalty term. For values between 1 and 2 a weighted sum of the difference penalization matrices is used.
rel_nsegis replaced by
knot_distance( distance between knots). Defaults to 0.025, which corresponds to the grid steps when knot_distance_power = 1 (the default).
knot_distance_powerdefines if knots are uniformly distributed. Defaults to 1, which corresponds to the equidistant case. Values less than 1 create more knots in the center, while values above 1 concentrate more knots in the tails.
allow_quantile_crossingdefines if quantile crossing is allowed. Defaults to false, which means that predictions will be sorted.
ymust now be a matrix of either T × 1 or T × P.
tracespecifies whether a progress bar will be printed or not. Default to
loss_functionlets you now specify “quantile”, “expectile” or “percentage”. All functions are generalized as in Gneitling 2009. The power can be scaled by
loss_parameter. The latter defaults to 1, which leads to the well-known quantile, squared, and absolute percentage loss.
gradientlets you specify whether the learning algorithm should consider actual loss or a linearized version using the gradient of the loss. Defaults to
forget_performancewas added. It defines the share of the past performance that will be ignored when selecting the best parameter combination.
forget_regretto underline its reference to the regret.
init_weightsparameter. It has to be either a Kx1 or KxP matrix specifying the experts’ starting weights.
lead_timeparameter. offset for expert forecasts. Defaults to 0 which means that experts predict t+1 at t. Setting this to h means experts’ predictions refer to t+1+h at time t. The weight updates delay accordingly.
tauis now optional. It defaults to 1:P/(P+1). A scalar given to tau will be repeated P times. The latter is useful in multivariate settings.
weightsobject is changed from a (T + 1 × K × P) array to a (T + 1 × P × K) array to match other objects’ dimensions. Now the following indexing scheme is consistent throughout the package: (Time, Probabilities, Experts, Parameter combination)