- ``lambdarank``, `lambdarank <https://proceedings.neurips.cc/paper_files/paper/2006/file/af44c4c56f385c43f2529f9b1b018f6a-Paper.pdf>`__ objective. `label_gain <#label_gain>`__ can be used to set the gain (weight) of ``int`` label and all values in ``label`` must be smaller than number of elements in ``label_gain``
- ``lambdarank``, `lambdarank <https://proceedings.neurips.cc/paper/2006/hash/af44c4c56f385c43f2529f9b1b018f6a-Abstract.html>`__ objective. `label_gain <#label_gain>`__ can be used to set the gain (weight) of ``int`` label and all values in ``label`` must be smaller than number of elements in ``label_gain``
- ``linear_lambda`` :raw-html:`<a id="linear_lambda" title="Permalink to this parameter" href="#linear_lambda">🔗︎</a>`, default = ``0.0``, type = double, constraints: ``linear_lambda >= 0.0``
- linear tree regularization, corresponds to the parameter ``lambda`` in Eq. 3 of `Gradient Boosting with Piece-Wise Linear Regression Trees <https://arxiv.org/pdf/1802.05640.pdf>`__
- linear tree regularization, corresponds to the parameter ``lambda`` in Eq. 3 of `Gradient Boosting with Piece-Wise Linear Regression Trees <https://arxiv.org/abs/1802.05640>`__
- ``min_gain_to_split`` :raw-html:`<a id="min_gain_to_split" title="Permalink to this parameter" href="#min_gain_to_split">🔗︎</a>`, default = ``0.0``, type = double, aliases: ``min_split_gain``, constraints: ``min_gain_to_split >= 0.0``
...
...
@@ -845,7 +845,7 @@ Dataset Parameters
- ``enable_bundle`` :raw-html:`<a id="enable_bundle" title="Permalink to this parameter" href="#enable_bundle">🔗︎</a>`, default = ``true``, type = bool, aliases: ``is_enable_bundle``, ``bundle``
- set this to ``false`` to disable Exclusive Feature Bundling (EFB), which is described in `LightGBM: A Highly Efficient Gradient Boosting Decision Tree <https://papers.nips.cc/paper_files/paper/2017/hash/6449f44a102fde848669bdd9eb6b76fa-Abstract.html>`__
- set this to ``false`` to disable Exclusive Feature Bundling (EFB), which is described in `LightGBM: A Highly Efficient Gradient Boosting Decision Tree <https://proceedings.neurips.cc/paper/2017/hash/6449f44a102fde848669bdd9eb6b76fa-Abstract.html>`__
- **Note**: disabling this may cause the slow training speed for sparse datasets
...
...
@@ -1192,7 +1192,7 @@ Objective Parameters
- used only in ``lambdarank`` application
- controls the number of top-results to focus on during training, refer to "truncation level" in the Sec. 3 of `LambdaMART paper <https://www.microsoft.com/en-us/research/wp-content/uploads/2016/02/MSR-TR-2010-82.pdf>`__
- controls the number of top-results to focus on during training, refer to "truncation level" in the Sec. 3 of `LambdaMART paper <https://www.microsoft.com/en-us/research/publication/from-ranknet-to-lambdarank-to-lambdamart-an-overview/>`__
- this parameter is closely related to the desirable cutoff ``k`` in the metric **NDCG@k** that we aim at optimizing the ranker for. The optimal setting for this parameter is likely to be slightly higher than ``k`` (e.g., ``k + 3``) to include more pairs of documents to train on, but perhaps not too high to avoid deviating too much from the desired target metric **NDCG@k**
* - param is only used by the CLI (especially the "predict" and "convert_model" tasks)
* - param is related to LightGBM writing files (e.g. "output_model", "save_binary")
*/
#ifndef LIGHTGBM_CONFIG_H_
#define LIGHTGBM_CONFIG_H_
#ifndef LIGHTGBM_INCLUDE_LIGHTGBM_CONFIG_H_
#define LIGHTGBM_INCLUDE_LIGHTGBM_CONFIG_H_
#include <LightGBM/export.h>
#include <LightGBM/meta.h>
...
...
@@ -154,7 +154,7 @@ struct Config {
// descl2 = ``cross_entropy_lambda``, alternative parameterization of cross-entropy, aliases: ``xentlambda``
// descl2 = label is anything in interval [0, 1]
// desc = ranking application
// descl2 = ``lambdarank``, `lambdarank <https://proceedings.neurips.cc/paper_files/paper/2006/file/af44c4c56f385c43f2529f9b1b018f6a-Paper.pdf>`__ objective. `label_gain <#label_gain>`__ can be used to set the gain (weight) of ``int`` label and all values in ``label`` must be smaller than number of elements in ``label_gain``
// descl2 = ``lambdarank``, `lambdarank <https://proceedings.neurips.cc/paper/2006/hash/af44c4c56f385c43f2529f9b1b018f6a-Abstract.html>`__ objective. `label_gain <#label_gain>`__ can be used to set the gain (weight) of ``int`` label and all values in ``label`` must be smaller than number of elements in ``label_gain``
// descl2 = ``rank_xendcg`` is faster than and achieves the similar performance as ``lambdarank``
// descl2 = label should be ``int`` type, and larger number represents the higher relevance (e.g. 0:bad, 1:fair, 2:good, 3:perfect)
...
...
@@ -423,7 +423,7 @@ struct Config {
doublelambda_l2=0.0;
// check = >=0.0
// desc = linear tree regularization, corresponds to the parameter ``lambda`` in Eq. 3 of `Gradient Boosting with Piece-Wise Linear Regression Trees <https://arxiv.org/pdf/1802.05640.pdf>`__
// desc = linear tree regularization, corresponds to the parameter ``lambda`` in Eq. 3 of `Gradient Boosting with Piece-Wise Linear Regression Trees <https://arxiv.org/abs/1802.05640>`__
doublelinear_lambda=0.0;
// alias = min_split_gain
...
...
@@ -706,7 +706,7 @@ struct Config {
boolis_enable_sparse=true;
// alias = is_enable_bundle, bundle
// desc = set this to ``false`` to disable Exclusive Feature Bundling (EFB), which is described in `LightGBM: A Highly Efficient Gradient Boosting Decision Tree <https://papers.nips.cc/paper_files/paper/2017/hash/6449f44a102fde848669bdd9eb6b76fa-Abstract.html>`__
// desc = set this to ``false`` to disable Exclusive Feature Bundling (EFB), which is described in `LightGBM: A Highly Efficient Gradient Boosting Decision Tree <https://proceedings.neurips.cc/paper/2017/hash/6449f44a102fde848669bdd9eb6b76fa-Abstract.html>`__
// desc = **Note**: disabling this may cause the slow training speed for sparse datasets
boolenable_bundle=true;
...
...
@@ -977,7 +977,7 @@ struct Config {
// check = >0
// desc = used only in ``lambdarank`` application
// desc = controls the number of top-results to focus on during training, refer to "truncation level" in the Sec. 3 of `LambdaMART paper <https://www.microsoft.com/en-us/research/wp-content/uploads/2016/02/MSR-TR-2010-82.pdf>`__
// desc = controls the number of top-results to focus on during training, refer to "truncation level" in the Sec. 3 of `LambdaMART paper <https://www.microsoft.com/en-us/research/publication/from-ranknet-to-lambdarank-to-lambdamart-an-overview/>`__
// desc = this parameter is closely related to the desirable cutoff ``k`` in the metric **NDCG@k** that we aim at optimizing the ranker for. The optimal setting for this parameter is likely to be slightly higher than ``k`` (e.g., ``k + 3``) to include more pairs of documents to train on, but perhaps not too high to avoid deviating too much from the desired target metric **NDCG@k**
