quantlab.portfolio.predefined.approximate.topk_dropout.decide#

quantlab.portfolio.predefined.approximate.topk_dropout.decide(rule, panel, prices=None, stopping=Stopping(tolerance=1e-05, max_iterations=20000))[source]#

Return every bar’s TopkDropout book, the bars decided forward from prices.previous.

Parameters:

ruleTopkDropoutConstructor: The rule: its config’s direction, topk and n_drop, and the label it ranks by (ranking_label).
panelDecisionPanel: The bars’ predictions and tradability.
pricesPanelPrices, optional: The book before the bars and the prices it drifts with; without them the book starts empty and does not drift.
stoppingStopping: Not read.

Returns:

PanelSolution: [bars, symbols] weights, no iterations, every bar converged.

Examples

>>> import numpy as np, pandas as pd, xarray as xr
>>> from quantlab.portfolio.base import DecisionPanel
>>> from quantlab.portfolio.config import TopkDropoutConfig
>>> from quantlab.portfolio.predefined.topk_dropout import TopkDropoutConstructor
>>> bars, symbols = pd.DatetimeIndex(["2024-01-02", "2024-01-03"]), ["AAA", "BBB", "CCC", "DDD"]
>>> panel = DecisionPanel(
...     timestamps=bars,
...     symbols=np.array(symbols),
...     predictions=xr.Dataset(
...         {"ret": (("timestamp", "symbol"), [[0.3, 0.2, 0.1, 0.0], [0.0, 0.2, 0.1, 0.4]])},
...         coords={"timestamp": bars, "symbol": symbols},
...     ),
...     tradable=np.ones((2, 4), dtype=bool),
... )
>>> decide(TopkDropoutConstructor(TopkDropoutConfig("long_only", topk=2, n_drop=1)), panel).weights
array([[0.5, 0.5, 0. , 0. ],
       [0. , 0.5, 0. , 0.5]])