quantlab-factor-agent 0.1.0__py3-none-any.whl

factor_agent/KUNQUANT_LICENSE

@@ -0,0 +1,201 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

factor_agent/__init__.py

@@ -0,0 +1,8 @@
+"""LLM-driven, mechanism-first formulaic factor research on quantlab."""
+
+from factor_agent import _expression_fold, _max_drawdown
+
+__version__ = "0.1.0"
+
+_expression_fold.patch()
+_max_drawdown.patch()

factor_agent/_acceptance_model.py

@@ -0,0 +1,93 @@
+"""The acceptance model: a quantlab model template that Acceptance fills in and trains.
+
+``acceptance.model`` is the component config (a ``get_config()`` dict) of a
+quantlab ``BaseModel``, or of a ``ModelEnsemble`` whose ``members`` are such
+configs. A model's ``factors`` are the baseline: entries as the campaign
+config's ``reference`` lists them, or empty (or absent) for the reference set
+itself. Everything else of the template, its ``hyperparameters`` above all,
+is the model's own.
+
+``build`` fills the template in: each model (each member of an ensemble) gets
+its baseline followed by the factors added, and the fields Acceptance owns
+(``FILLED``: the label, the save directory, the read strategies, the
+windows, ``val_size``, the scored split and the seed) replace whatever the
+template says of them, so a config copied from a trained run's
+``config.json`` is a template as it is. A ``SeedEnsemble`` template is
+refused: Acceptance trains every model once per ``acceptance.seeds``.
+"""
+
+import dataclasses
+from collections.abc import Callable
+
+from quantlab.core.component import get_cls_from_path, rebuild
+from quantlab.model.base import BaseModel
+from quantlab.model.config import ModelConfig
+from quantlab.model.predefined.model_ensemble import ModelEnsemble
+from quantlab.model.predefined.seed_ensemble import SeedEnsemble
+
+#: The ``ModelConfig`` fields Acceptance sets in every model, whatever the template says.
+FILLED = (
+    "labels", "model_save_dir", "factor_data_strategy", "label_data_strategy", "start_date", "end_date",
+    "train_start", "train_end", "test_start", "test_end", "val_size", "metric_splits", "random_seed",
+)
+#: The ``ModelConfig`` fields a template may give.
+_FIELDS = {f.name for f in dataclasses.fields(ModelConfig)}
+
+
+def check(template: dict) -> None:
+    """Refuse a template that ``build`` cannot fill in, before anything is read or trained.
+
+    Raises
+    ------
+    ValueError
+        If the template is a ``SeedEnsemble`` (seeds are ``acceptance.seeds``).
+    TypeError
+        If it names no class, a class that is neither a ``BaseModel`` nor a
+        ``ModelEnsemble``, or a model with a key ``ModelConfig`` lacks.
+    """
+    if not isinstance(template, dict) or "name" not in template:
+        raise TypeError("acceptance.model: the model's component config needs its class path as name")
+    cls = get_cls_from_path(template["name"])
+    if not isinstance(cls, type):
+        raise TypeError(f"acceptance.model: {template['name']} is not a BaseModel or a ModelEnsemble")
+    if issubclass(cls, SeedEnsemble):
+        raise ValueError(
+            "acceptance.model: a SeedEnsemble is not a template; give its model and set the seeds as acceptance.seeds"
+        )
+    if issubclass(cls, ModelEnsemble):
+        unknown = set(template) - {"name", "members"}
+        if unknown:
+            raise TypeError(f"acceptance.model: the ModelEnsemble has unknown key(s) {sorted(unknown)}")
+        for member in template.get("members") or ():
+            check(member)
+        return
+    if not issubclass(cls, BaseModel):
+        raise TypeError(f"acceptance.model: {template['name']} is not a BaseModel or a ModelEnsemble")
+    unknown = set(template) - _FIELDS
+    if unknown:
+        raise TypeError(f"acceptance.model: {template['name']} has key(s) {sorted(unknown)} ModelConfig lacks")
+
+
+def build(template: dict, baseline: Callable[[tuple], list], added: list, filled: dict):
+    """The model (or ``ModelEnsemble``) of ``template``, filled in and untrained.
+
+    Parameters
+    ----------
+    template : dict
+        ``acceptance.model``; see the module docstring.
+    baseline : callable
+        A model's ``factors`` entries (empty for the reference set) to its baseline features.
+    added : list
+        The features added to every model after its baseline.
+    filled : dict
+        A value for each of ``FILLED``, set in every model.
+    """
+    check(template)
+    cls = get_cls_from_path(template["name"])
+    if issubclass(cls, ModelEnsemble):
+        return cls([build(member, baseline, added, filled) for member in template.get("members") or ()])
+    fields = {key: value for key, value in template.items() if key not in ("name", *FILLED)}
+    if "tracker" in fields:
+        fields["tracker"] = rebuild(fields["tracker"])
+    fields["factors"] = [*baseline(tuple(fields.get("factors") or ())), *added]
+    return cls(ModelConfig(**fields, **filled))

factor_agent/_admission.py

@@ -0,0 +1,168 @@
+"""The admission policy: a pure function of a candidate's metrics and the trial count.
+
+A candidate states no direction: its direction is the sign of its training
+IC (``direction``), and the statistics below are read in it. The checks run
+in order and the first that fails is the reason:
+
+1. on an average training day the candidate has a value on at least
+   ``min_coverage`` of the members (its other statistics speak only for
+   the members it covers); the campaign adds to the refusal where the
+   graph loses its values (``coverage_cause``);
+2. its training IC is a nonzero number, so it has a direction;
+3. its correlation with every reference and accepted factor is at most
+   ``max_correlation`` in absolute value, and so is its sampled correlation
+   with every known factor (on every ``KNOWN_SAMPLE_EVERY``-th scored date
+   of the Evaluator);
+4. its partial IC, in its direction, has a Newey-West t at or above the
+   threshold for the number of evaluated trials so far;
+5. its decile returns, in its direction, rise with the decile: their
+   Spearman correlation with the decile number is at least
+   ``min_monotonicity``;
+6. with a confirmation segment, its partial IC there, in its direction,
+   has a Newey-West t of at least ``confirm_t`` (``confirmed``); the
+   reason says so without a number, since no prompt shows a statistic of
+   that segment.
+
+The threshold is a Bonferroni correction anchored at ``t_threshold``
+(Harvey, Liu and Zhu 2016). The direction is read from the data, so each
+trial is two tests, one per direction: the one-sided p-value of
+``t_threshold`` is divided by twice the trial count n (3.0 becomes about
+3.21 at one trial, 3.82 at 10, 4.35 at 100 and 4.73 at 600). With
+``trial_exponent`` a below 1 it is divided by twice n to the a instead, a
+threshold rising more slowly than Bonferroni's (at 0.5, 3.82 at 100 and
+4.03 at 600).
+"""
+
+import math
+from collections.abc import Callable, Sequence
+
+import numpy as np
+from scipy import stats
+
+from factor_agent.config import AdmissionConfig
+from factor_agent.evaluator import KNOWN_SAMPLE_EVERY, Metrics, NodeCoverage
+
+#: The dates a known factor's correlation is sampled on, as the reason says.
+_SAMPLED_DATES = f"one training date in {KNOWN_SAMPLE_EVERY}"
+
+
+def t_threshold(config: AdmissionConfig, trials: int) -> float:
+    """The partial-IC Newey-West t a candidate needs as the ``trials``-th evaluated trial."""
+    return float(stats.norm.isf(stats.norm.sf(config.t_threshold) / (2 * max(trials, 1) ** config.trial_exponent)))
+
+
+def direction(ic_mean: float) -> int:
+    """A factor's direction: +1 when a higher value went with a higher training return, -1 when lower, 0 when neither."""
+    return int(np.sign(ic_mean)) if math.isfinite(ic_mean) else 0
+
+
+def monotonicity(decile_returns, sign: int) -> float:
+    """The Spearman correlation of the decile number with the decile returns, signed by ``sign``."""
+    returns = np.asarray(decile_returns, dtype=float)
+    if not np.isfinite(returns).all() or np.ptp(returns) == 0:
+        return float("nan")
+    return sign * float(stats.spearmanr(np.arange(len(returns)), returns).statistic)
+
+
+#: What each admission reason starts with, and the kind of refusal it is (``reason_kind``).
+_KINDS = (
+    ("coverage ", "coverage"),
+    ("the training IC ", "no direction"),
+    ("cross-sectional correlation ", "correlation"),
+    ("the partial IC's ", "partial IC"),
+    ("decile monotonicity ", "monotonicity"),
+    ("the factor did not hold on the confirmation period", "confirmation"),
+)
+
+#: Why a candidate that passed every other check is refused when its partial IC does not hold on the
+#: confirmation segment: without a number, since no prompt may show a statistic of that segment.
+NOT_CONFIRMED = (
+    "the factor did not hold on the confirmation period, a later part of the training data whose "
+    "statistics the search is not shown: its residual IC there was too weak or went the other way"
+)
+
+
+def confirmed(ic_mean: float, confirmation_partial_t: float, config: AdmissionConfig) -> bool:
+    """Whether the partial IC on the confirmation segment holds in the direction of the training ``ic_mean``."""
+    return direction(ic_mean) * confirmation_partial_t >= config.confirm_t
+
+
+def reason_kind(reason: str) -> str:
+    """The kind of an admission reason ``reasons`` wrote: coverage, no direction, correlation, partial IC or monotonicity."""
+    return next((kind for start, kind in _KINDS if reason.startswith(start)), "other")
+
+
+def coverage_cause(trace: Sequence[NodeCoverage], render: Callable[[int], str], min_coverage: float) -> str:
+    """Where a candidate refused for its coverage loses its values, from its ``trace`` (``Evaluator.trace_coverage``).
+
+    The step named is the first traced node below ``min_coverage`` whose
+    traced inputs are all at or above it, with the coverage of those inputs
+    and the share on which the step or an input is infinite; a data column
+    below it is named as the data it reads. ``render`` writes a node as
+    candidate code writes it. The text continues an admission reason; it is
+    empty when no such node is traced.
+    """
+    by_node = {t.node: t for t in trace}
+    for t in trace:
+        given = [by_node[j] for j in t.inputs]
+        if t.coverage >= min_coverage or any(g.coverage < min_coverage for g in given):
+            continue
+        if t.column:
+            return f"; the data column {render(t.node)} has a value on {t.coverage:.2f} of the member days"
+        cause = f"; it falls at {render(t.node)}, to {t.coverage:.2f}"
+        if given:
+            cause += " from " + ", ".join(f"{g.coverage:.2f} for {render(g.node)}" for g in given)
+        infinite = [(f"{render(g.node)} is", g) for g in given if g.infinite > 0]
+        if t.infinite > 0:
+            infinite = [("it is", t)]
+        for subject, g in infinite:
+            cause += (
+                f"; {subject} infinite on {_share(g.infinite)} of the member days and on {_share(g.infinite_days)} "
+                f"of the days at least once"
+            )
+        if infinite:
+            cause += " (a division by zero is infinite)"
+        return cause
+    return ""
+
+
+def _share(fraction: float) -> str:
+    """A share as a percentage, with a decimal below 1%, so a small nonzero share does not read 0%."""
+    return f"{fraction:.1%}" if fraction < 0.01 else f"{fraction:.0%}"
+
+
+def reasons(metrics: Metrics, trials: int, config: AdmissionConfig) -> list[str]:
+    """Why a candidate with ``metrics`` is refused, as one reason, or nothing when it is accepted."""
+    sign = direction(metrics.ic_mean)
+    if not metrics.coverage >= config.min_coverage:
+        return [
+            f"coverage {metrics.coverage:.2f} (the mean daily share of the members with a value) "
+            f"is below {config.min_coverage:g}"
+        ]
+    if sign == 0:
+        return [f"the training IC {metrics.ic_mean:+.4f} gives the factor no direction"]
+    if abs(metrics.nearest_correlation) > config.max_correlation:
+        return [
+            f"cross-sectional correlation {metrics.nearest_correlation:+.2f} with {metrics.nearest} "
+            f"is above {config.max_correlation:g} in absolute value"
+        ]
+    if abs(metrics.nearest_known_correlation) > config.max_correlation:
+        return [
+            f"cross-sectional correlation {metrics.nearest_known_correlation:+.2f} with {metrics.nearest_known}, "
+            f"sampled on {_SAMPLED_DATES}, is above {config.max_correlation:g} in absolute value"
+        ]
+    threshold = t_threshold(config, trials)
+    partial_t = sign * metrics.partial_ic_t
+    if not partial_t >= threshold:
+        return [
+            f"the partial IC's Newey-West t {partial_t:.2f} (residualised against the reference set and "
+            f"the accepted factors, in the factor's direction) is below the threshold {threshold:.2f} "
+            f"at {trials} evaluated trial(s), each counted in both directions"
+        ]
+    rising = monotonicity(metrics.decile_returns, sign)
+    if not rising >= config.min_monotonicity:
+        return [
+            f"decile monotonicity {rising:.2f} (Spearman correlation of the decile returns with the decile, "
+            f"in the factor's direction) is below {config.min_monotonicity:g}"
+        ]
+    return []

factor_agent/_expression_fold.py

@@ -0,0 +1,48 @@
+"""KunQuant's expression folding, with operators compared as graphs.
+
+KunQuant's ``expr_fold`` pass merges equal operators. Two whose hashes match
+and whose inputs are not the same objects it tells apart by walking both as
+trees (``OpBase.items``), which visits a value once for every path to it: a
+graph that reuses its values, as ``x / Scale(x)`` reuses ``x``, nested a few
+times under an EW beta, takes longer to compare than any round can wait (an
+evaluation of arm A2's round 6 ran for hours). ``patch`` makes the comparison
+the same walk over the graph, side by side and with a stack, visiting each pair
+of operators once; it is the fix proposed to KunQuant
+(https://github.com/Menooker/KunQuant/pull/83), and a KunQuant that has it is
+left as it is.
+"""
+
+from KunQuant.passes import ExprFold
+
+
+def _same(x, y) -> bool:
+    """Whether ``x.items()`` and ``y.items()`` are equal."""
+    stack = [(x, {}, y, {})]
+    visited = set()
+    while stack:
+        a, a_args, b, b_args = stack.pop()
+        if a is b and a_args.keys() == b_args.keys() and all(a_args[k] is b_args[k] for k in a_args):
+            continue
+        key = (id(a), id(b), *((k, id(v)) for k, v in a_args.items()), *((k, id(v)) for k, v in b_args.items()))
+        if key in visited:
+            continue
+        visited.add(key)
+        if a.__class__ is not b.__class__ or a.attrs != b.attrs:
+            return False
+        a_inputs, b_inputs = a.get_args(True, **a_args), b.get_args(True, **b_args)
+        if len(a_inputs) != len(b_inputs):
+            return False
+        stack += [(c, c_args, d, d_args) for (c, c_args), (d, d_args) in zip(a_inputs, b_inputs)]
+    return True
+
+
+def _equal(self, other) -> bool:
+    if self.parent != other.parent or self.op.__class__ != other.op.__class__ or self.op.attrs != other.op.attrs:
+        return False
+    return self.op.inputs == other.op.inputs or _same(self.op, other.op)
+
+
+def patch() -> None:
+    """Make KunQuant's expression folding compare operators as graphs, unless it already does; calling it again does nothing."""
+    if not hasattr(ExprFold, "_is_same"):
+        ExprFold._OpProxy.__eq__ = _equal