News for 2.0. [skip ci] #9484

trivialfis · 2023-08-15T10:20:51Z

No description provided.

NEWS.md

trivialfis · 2023-08-19T16:12:07Z

NEWS.md

+In the previous version, `base_score` was a constant that could be set as a training parameter. In the new version, XGBoost can automatically estimate this parameter based on input labels for optimal accuracy. (#8539, #8498, #8272, #8793, #8607)
+
+### Quantile regression
+The XGBoost algorithm now supports quantile regression, which involves minimizing the quantile loss (also called "pinball loss"). Furthermore, XGBoost allows for training with multiple target quantiles simultaneously. (#8775, #8761, #8760, #8758, #8750)


Suggested change

The XGBoost algorithm now supports quantile regression, which involves minimizing the quantile loss (also called "pinball loss"). Furthermore, XGBoost allows for training with multiple target quantiles simultaneously. (#8775, #8761, #8760, #8758, #8750)

The XGBoost algorithm now supports quantile regression, which involves minimizing the quantile loss (also called "pinball loss"). Furthermore, XGBoost allows for training with multiple target quantiles simultaneously using one tree per quantile. (#8775, #8761, #8760, #8758, #8750)

When we fit a tree per quantile, is the resulting model considered a multi-target model?

NEWS.md

trivialfis · 2023-08-19T16:19:42Z

NEWS.md

+
+* doc
+- Add introduction and notes for the sklearn interface. (#8948)
+- Demo for using dask with HPO. (#8891)


Suggested change

- Demo for using dask with HPO. (#8891)

- Demo for using dask for HPO. (#8891)

Suggested change

- Demo for using dask with HPO. (#8891)

- Demo for using dask for hyperparameter optimization (HPO). (#8891)

trivialfis · 2023-08-19T16:20:56Z

NEWS.md

+* CI bot PRs
+We employed GitHub dependent bot to help us keep the dependencies up-to-date for JVM packages. With the help from the bot, we have cleared up all the dependencies that are lagging behind (#8501, #8507).
+
+Here's a list of dependency update PRs including those made by dependent bots (#8456, #8560, #8571, #8561, #8562, #8600, #8594, #8524, #8509, #8548, #8549, #8533, #8521, #8534, #8532, #8516, #8503, #8531, #8530, #8518, #8512, #8515, #8517, #8506, #8504, #8502, #8629, #8815, #8813, #8814, #8877, #8876, #8875, #8874, #8873, #9049, #9070, #9073, #9039, #9083, #8917, #8952, #8980, #8973, #8962, #9252, #9208, #9131, #9136, #9219, #9160, #9158, #9163, #9184, #9192, #9265, #9268, #8882, #8837, #8662, #8661, #8390, #9056, #8508, #8925, #8920, #9149, #9230, #9097, #8648, #9203, 8593).


Suggested change

Here's a list of dependency update PRs including those made by dependent bots (#8456, #8560, #8571, #8561, #8562, #8600, #8594, #8524, #8509, #8548, #8549, #8533, #8521, #8534, #8532, #8516, #8503, #8531, #8530, #8518, #8512, #8515, #8517, #8506, #8504, #8502, #8629, #8815, #8813, #8814, #8877, #8876, #8875, #8874, #8873, #9049, #9070, #9073, #9039, #9083, #8917, #8952, #8980, #8973, #8962, #9252, #9208, #9131, #9136, #9219, #9160, #9158, #9163, #9184, #9192, #9265, #9268, #8882, #8837, #8662, #8661, #8390, #9056, #8508, #8925, #8920, #9149, #9230, #9097, #8648, #9203, 8593).

Here's a list of dependency update PRs including those made by dependent bots (#8456, #8560, #8571, #8561, #8562, #8600, #8594, #8524, #8509, #8548, #8549, #8533, #8521, #8534, #8532, #8516, #8503, #8531, #8530, #8518, #8512, #8515, #8517, #8506, #8504, #8502, #8629, #8815, #8813, #8814, #8877, #8876, #8875, #8874, #8873, #9049, #9070, #9073, #9039, #9083, #8917, #8952, #8980, #8973, #8962, #9252, #9208, #9131, #9136, #9219, #9160, #9158, #9163, #9184, #9192, #9265, #9268, #8882, #8837, #8662, #8661, #8390, #9056, #8508, #8925, #8920, #9149, #9230, #9097, #8648, #9203, #8593).

hcho3 · 2023-08-30T20:50:01Z

NEWS.md

+### Initial work on multi-target tree
+We have been working on vector-leaf tree models for multi-target regression, multi-label classification, and multi-class classification in version 2.0. Previously, XGBoost would build a separate model for each target. However, with this new feature that's still being developed, XGBoost can create one tree for all targets. This method has various benefits and trade-offs compared to the current approach. It can help prevent overfitting, result in smaller models, and build trees that consider the correlation between targets. Additionally, users can combine vector leaf and scalar leaf trees during a training session using a callback. Please note that this feature is still a work in progress, and many parts are not yet available. See #9043 for the current status. Related PRs: (#8538, #8697, #8902, #8884, #8895, #8898, #8612, #8652, #8698, #8908, #8928, #8968, #8616, #8922, #8890, #8872, #8889) Please note that only the `hist` (default) tree method on CPU can be used for building vector leaf trees at the moment.


Suggested change

### Initial work on multi-target tree

We have been working on vector-leaf tree models for multi-target regression, multi-label classification, and multi-class classification in version 2.0. Previously, XGBoost would build a separate model for each target. However, with this new feature that's still being developed, XGBoost can create one tree for all targets. This method has various benefits and trade-offs compared to the current approach. It can help prevent overfitting, result in smaller models, and build trees that consider the correlation between targets. Additionally, users can combine vector leaf and scalar leaf trees during a training session using a callback. Please note that this feature is still a work in progress, and many parts are not yet available. See #9043 for the current status. Related PRs: (#8538, #8697, #8902, #8884, #8895, #8898, #8612, #8652, #8698, #8908, #8928, #8968, #8616, #8922, #8890, #8872, #8889) Please note that only the `hist` (default) tree method on CPU can be used for building vector leaf trees at the moment.

### Initial work on multi-target trees with vector-leaf outputs

We have been working on vector-leaf tree models for multi-target regression, multi-label classification, and multi-class classification in version 2.0. Previously, XGBoost would build a separate model for each target. In 2.0, users can optionally fit trees with vector-leaf outputs, so that one tree yields prediction for all targets. The vector-leaf tree method has various benefits and trade-offs compared to the current approach. It can help prevent overfitting, result in smaller models, and build trees that consider correlations between targets. Additionally, users can combine vector-leaf and scalar-leaf trees during a training session using a callback. Please note that this feature is still a work in progress, and many parts are not yet available. See #9043 for the current status. Related PRs: (#8538, #8697, #8902, #8884, #8895, #8898, #8612, #8652, #8698, #8908, #8928, #8968, #8616, #8922, #8890, #8872, #8889) Please note that only the `hist` (default) tree method on CPU can be used for building vector-leaf trees at the moment.

NEWS.md

hcho3 · 2023-08-30T21:36:47Z

NEWS.md

+
+* doc
+- Add introduction and notes for the sklearn interface. (#8948)
+- Demo for using dask with HPO. (#8891)


Suggested change

- Demo for using dask with HPO. (#8891)

- Demo for using dask for hyperparameter optimization (HPO). (#8891)

NEWS.md

trivialfis · 2023-09-18T20:14:34Z

cc @hcho3 .

trivialfis added 2 commits August 15, 2023 18:20

[WIP] news for 2.0. [skip ci]

783d2ba

update [skip ci]

8813359

trivialfis added this to 2.0 In Progress in 2.0 Roadmap via automation Aug 15, 2023

trivialfis mentioned this pull request Aug 17, 2023

2.0.0 Release Candidate #9497

Closed

5 tasks

wording, typos. [skip ci]

185eff5

trivialfis commented Aug 17, 2023

View reviewed changes

NEWS.md Outdated Show resolved Hide resolved

trivialfis commented Aug 17, 2023

View reviewed changes

NEWS.md Outdated Show resolved Hide resolved

trivialfis commented Aug 19, 2023

View reviewed changes

hcho3 requested changes Aug 30, 2023

View reviewed changes

trivialfis added 3 commits September 12, 2023 10:06

Update with backported PRs.

160b865

reviewer's comments. [skip ci]

a6874ee

Merge remote-tracking branch 'jiamingy/2.0-news' into 2.0-news [skip ci]

3d68545

trivialfis changed the title ~~[WIP] news for 2.0. [skip ci]~~ News for 2.0. [skip ci] Sep 18, 2023

hcho3 approved these changes Sep 18, 2023

View reviewed changes

Merge remote-tracking branch 'upstream/master' into 2.0-news

5ffa51c

trivialfis merged commit 259d80c into dmlc:master Sep 18, 2023
22 checks passed

2.0 Roadmap automation moved this from 2.0 In Progress to 2.0 Done Sep 18, 2023

trivialfis deleted the 2.0-news branch September 18, 2023 21:46

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

News for 2.0. [skip ci] #9484

News for 2.0. [skip ci] #9484

trivialfis commented Aug 15, 2023

trivialfis Aug 19, 2023

hcho3 Aug 30, 2023

trivialfis Sep 12, 2023

trivialfis Aug 19, 2023

hcho3 Aug 30, 2023

trivialfis Aug 19, 2023

hcho3 Aug 30, 2023

hcho3 Aug 30, 2023

trivialfis commented Sep 18, 2023

	The XGBoost algorithm now supports quantile regression, which involves minimizing the quantile loss (also called "pinball loss"). Furthermore, XGBoost allows for training with multiple target quantiles simultaneously. (#8775, #8761, #8760, #8758, #8750)
	The XGBoost algorithm now supports quantile regression, which involves minimizing the quantile loss (also called "pinball loss"). Furthermore, XGBoost allows for training with multiple target quantiles simultaneously using one tree per quantile. (#8775, #8761, #8760, #8758, #8750)

	- Demo for using dask with HPO. (#8891)
	- Demo for using dask for HPO. (#8891)

	- Demo for using dask with HPO. (#8891)
	- Demo for using dask for hyperparameter optimization (HPO). (#8891)

		### Initial work on multi-target tree
		We have been working on vector-leaf tree models for multi-target regression, multi-label classification, and multi-class classification in version 2.0. Previously, XGBoost would build a separate model for each target. However, with this new feature that's still being developed, XGBoost can create one tree for all targets. This method has various benefits and trade-offs compared to the current approach. It can help prevent overfitting, result in smaller models, and build trees that consider the correlation between targets. Additionally, users can combine vector leaf and scalar leaf trees during a training session using a callback. Please note that this feature is still a work in progress, and many parts are not yet available. See #9043 for the current status. Related PRs: (#8538, #8697, #8902, #8884, #8895, #8898, #8612, #8652, #8698, #8908, #8928, #8968, #8616, #8922, #8890, #8872, #8889) Please note that only the `hist` (default) tree method on CPU can be used for building vector leaf trees at the moment.

News for 2.0. [skip ci] #9484

News for 2.0. [skip ci] #9484

Conversation

trivialfis commented Aug 15, 2023

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

trivialfis commented Sep 18, 2023