random forests and multil-class probability
Hello community, Do I understand correctly that Random Forests are trained as a 1 vs rest when the target has more than 2 classes? Say the target takes values 0, 1 and 2, then the model would train 3 estimators 1 per class under the hood?. The predict_proba output is an array with 3 columns, containing the probability of each class. If it is 1 vs rest. am I correct to assume that the sum of the probabilities for the 3 classes should not necessarily add up to 1? are they normalized? how is it done so that they do add up to 1? Thank you Sole
On 27 Jul 2021, at 11:08, Sole Galli via scikit-learn <scikit-learn@python.org> wrote:
Hello community,
Do I understand correctly that Random Forests are trained as a 1 vs rest when the target has more than 2 classes? Say the target takes values 0, 1 and 2, then the model would train 3 estimators 1 per class under the hood?.
Each decision tree of the forest is natively supporting multi class.
The predict_proba output is an array with 3 columns, containing the probability of each class. If it is 1 vs rest. am I correct to assume that the sum of the probabilities for the 3 classes should not necessarily add up to 1? are they normalized? how is it done so that they do add up to 1?
According to the above answer, the sum for each row of the array given by `predict_proba` will sum to 1. According to the documentation, the probabilities are computed as: The predicted class probabilities of an input sample are computed as the mean predicted class probabilities of the trees in the forest. The class probability of a single tree is the fraction of samples of the same class in a leaf.
Thank you Sole
_______________________________________________ scikit-learn mailing list scikit-learn@python.org https://mail.python.org/mailman/listinfo/scikit-learn
Thank you! I was confused because in the multiclass documentation it says that for those estimators that have multiclass support built in, like Decision trees and Random Forests, then we do not need to use the wrapper classes like the OnevsRest. Thus I have the following question, if I want to determine the PR curves or the ROC curve, say with micro-average, do I need to wrap them with the 1 vs rest? Or it does not matter? The probability values do change slightly. Thank you! ‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐ On Tuesday, July 27th, 2021 at 11:22 AM, Guillaume Lemaître <g.lemaitre58@gmail.com> wrote:
On 27 Jul 2021, at 11:08, Sole Galli via scikit-learn scikit-learn@python.org wrote:
Hello community,
Do I understand correctly that Random Forests are trained as a 1 vs rest when the target has more than 2 classes? Say the target takes values 0, 1 and 2, then the model would train 3 estimators 1 per class under the hood?.
Each decision tree of the forest is natively supporting multi class.
The predict_proba output is an array with 3 columns, containing the probability of each class. If it is 1 vs rest. am I correct to assume that the sum of the probabilities for the 3 classes should not necessarily add up to 1? are they normalized? how is it done so that they do add up to 1?
According to the above answer, the sum for each row of the array given by `predict_proba` will sum to 1.
According to the documentation, the probabilities are computed as:
The predicted class probabilities of an input sample are computed as the mean predicted class probabilities of the trees in the forest. The class probability of a single tree is the fraction of samples of the same class in a leaf.
Thank you
Sole
scikit-learn mailing list
scikit-learn@python.org
Greetings! I am currently out of office, with limited access to emails, till August the 30th. Please contact support@giotto.ai for technical issue concerning Giotto Platform. Otherwise, I will reply to your email as soon as possible upon my return. With best regards, Matteo On 27 Jul 2021, at 11:31, Sole Galli via scikit-learn <scikit-learn@python.org> wrote: Thank you! I was confused because in the multiclass documentation it says that for those estimators that have multiclass support built in, like Decision trees and Random Forests, then we do not need to use the wrapper classes like the OnevsRest. Thus I have the following question, if I want to determine the PR curves or the ROC curve, say with micro-average, do I need to wrap them with the 1 vs rest? Or it does not matter? The probability values do change slightly. Thank you! ‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐ On Tuesday, July 27th, 2021 at 11:22 AM, Guillaume Lemaître <g.lemaitre58@gmail.com> wrote: On 27 Jul 2021, at 11:08, Sole Galli via scikit-learn scikit-learn@python.org wrote: Hello community, Do I understand correctly that Random Forests are trained as a 1 vs rest when the target has more than 2 classes? Say the target takes values 0, 1 and 2, then the model would train 3 estimators 1 per class under the hood?. Each decision tree of the forest is natively supporting multi class. The predict_proba output is an array with 3 columns, containing the probability of each class. If it is 1 vs rest. am I correct to assume that the sum of the probabilities for the 3 classes should not necessarily add up to 1? are they normalized? how is it done so that they do add up to 1? According to the above answer, the sum for each row of the array given by `predict_proba` will sum to 1. According to the documentation, the probabilities are computed as: The predicted class probabilities of an input sample are computed as the mean predicted class probabilities of the trees in the forest. The class probability of a single tree is the fraction of samples of the same class in a leaf. Thank you Sole scikit-learn mailing list scikit-learn@python.org https://mail.python.org/mailman/listinfo/scikit-learn _______________________________________________ scikit-learn mailing list scikit-learn@python.org https://mail.python.org/mailman/listinfo/scikit-learn
Greetings! I am currently out of office, with limited access to emails, till August the 30th. Please contact support@giotto.ai for technical issues concerning Giotto Platform. Otherwise, I will reply to your email as soon as possible upon my return. With best regards, Matteo On 27 Jul 2021, at 11:31, Sole Galli via scikit-learn <scikit-learn@python.org> wrote: Thank you! I was confused because in the multiclass documentation it says that for those estimators that have multiclass support built in, like Decision trees and Random Forests, then we do not need to use the wrapper classes like the OnevsRest. Thus I have the following question, if I want to determine the PR curves or the ROC curve, say with micro-average, do I need to wrap them with the 1 vs rest? Or it does not matter? The probability values do change slightly. Thank you! ‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐ On Tuesday, July 27th, 2021 at 11:22 AM, Guillaume Lemaître <g.lemaitre58@gmail.com> wrote: On 27 Jul 2021, at 11:08, Sole Galli via scikit-learn scikit-learn@python.org wrote: Hello community, Do I understand correctly that Random Forests are trained as a 1 vs rest when the target has more than 2 classes? Say the target takes values 0, 1 and 2, then the model would train 3 estimators 1 per class under the hood?. Each decision tree of the forest is natively supporting multi class. The predict_proba output is an array with 3 columns, containing the probability of each class. If it is 1 vs rest. am I correct to assume that the sum of the probabilities for the 3 classes should not necessarily add up to 1? are they normalized? how is it done so that they do add up to 1? According to the above answer, the sum for each row of the array given by `predict_proba` will sum to 1. According to the documentation, the probabilities are computed as: The predicted class probabilities of an input sample are computed as the mean predicted class probabilities of the trees in the forest. The class probability of a single tree is the fraction of samples of the same class in a leaf. Thank you Sole scikit-learn mailing list scikit-learn@python.org https://mail.python.org/mailman/listinfo/scikit-learn _______________________________________________ scikit-learn mailing list scikit-learn@python.org https://mail.python.org/mailman/listinfo/scikit-learn
To add to Guillaume's answer: the native multiclass support for forests/trees is described here: https://scikit-learn.org/stable/modules/tree.html#multi-output-problems It's not a one-vs-rest strategy and can be summed up as:
*
Store n output values in leaves, instead of 1;
*
Use splitting criteria that compute the average reduction across all n outputs.
Nicolas On 27/07/2021 10:22, Guillaume Lemaître wrote:
On 27 Jul 2021, at 11:08, Sole Galli via scikit-learn <scikit-learn@python.org> wrote:
Hello community,
Do I understand correctly that Random Forests are trained as a 1 vs rest when the target has more than 2 classes? Say the target takes values 0, 1 and 2, then the model would train 3 estimators 1 per class under the hood?. Each decision tree of the forest is natively supporting multi class.
The predict_proba output is an array with 3 columns, containing the probability of each class. If it is 1 vs rest. am I correct to assume that the sum of the probabilities for the 3 classes should not necessarily add up to 1? are they normalized? how is it done so that they do add up to 1? According to the above answer, the sum for each row of the array given by `predict_proba` will sum to 1. According to the documentation, the probabilities are computed as:
The predicted class probabilities of an input sample are computed as the mean predicted class probabilities of the trees in the forest. The class probability of a single tree is the fraction of samples of the same class in a leaf.
Thank you Sole
_______________________________________________ scikit-learn mailing list scikit-learn@python.org https://mail.python.org/mailman/listinfo/scikit-learn
scikit-learn mailing list scikit-learn@python.org https://mail.python.org/mailman/listinfo/scikit-learn
Thank you! So when in the multiclass document says that for the algorithms that support intrinsically multiclass, which are listed [here](https://scikit-learn.org/stable/modules/multiclass.html), when it says that they do not need to be wrapped by the OnevsRest, it means that there is no need, because they can indeed handle multi class, each one in their own way. But, if I want to plot PR curves or ROC curves, then I do need to wrap them because those metrics are calculated as a 1 vs rest manner, and this is not how it is handled by the algos. Is my understanding correct? Thank you! ‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐ On Tuesday, July 27th, 2021 at 11:33 AM, Nicolas Hug <niourf@gmail.com> wrote:
To add to Guillaume's answer: the native multiclass support for forests/trees is described here: https://scikit-learn.org/stable/modules/tree.html#multi-output-problems
It's not a one-vs-rest strategy and can be summed up as:
-
Store n output values in leaves, instead of 1;
-
Use splitting criteria that compute the average reduction across all n outputs.
Nicolas
On 27/07/2021 10:22, Guillaume Lemaître wrote:
On 27 Jul 2021, at 11:08, Sole Galli via scikit-learn [<scikit-learn@python.org>](mailto:scikit-learn@python.org) wrote:
Hello community,
Do I understand correctly that Random Forests are trained as a 1 vs rest when the target has more than 2 classes? Say the target takes values 0, 1 and 2, then the model would train 3 estimators 1 per class under the hood?.
Each decision tree of the forest is natively supporting multi class.
The predict_proba output is an array with 3 columns, containing the probability of each class. If it is 1 vs rest. am I correct to assume that the sum of the probabilities for the 3 classes should not necessarily add up to 1? are they normalized? how is it done so that they do add up to 1?
According to the above answer, the sum for each row of the array given by `predict_proba` will sum to 1. According to the documentation, the probabilities are computed as:
The predicted class probabilities of an input sample are computed as the mean predicted class probabilities of the trees in the forest. The class probability of a single tree is the fraction of samples of the same class in a leaf.
Thank you Sole
_______________________________________________ scikit-learn mailing list scikit-learn@python.org
_______________________________________________ scikit-learn mailing list scikit-learn@python.org
As far that I remember, `precision_recall_curve` and `roc_curve` do not support multi class. They are design to work only with binary classification. Then, we provide an example for precision-recall that shows one way to compute precision-recall curve via averaging: https://scikit-learn.org/stable/auto_examples/model_selection/plot_precision... <https://scikit-learn.org/stable/auto_examples/model_selection/plot_precision...> -- Guillaume Lemaitre Scikit-learn @ Inria Foundation https://glemaitre.github.io/
On 27 Jul 2021, at 11:42, Sole Galli via scikit-learn <scikit-learn@python.org> wrote:
Thank you!
So when in the multiclass document says that for the algorithms that support intrinsically multiclass, which are listed here <https://scikit-learn.org/stable/modules/multiclass.html>, when it says that they do not need to be wrapped by the OnevsRest, it means that there is no need, because they can indeed handle multi class, each one in their own way.
But, if I want to plot PR curves or ROC curves, then I do need to wrap them because those metrics are calculated as a 1 vs rest manner, and this is not how it is handled by the algos. Is my understanding correct?
Thank you!
‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐ On Tuesday, July 27th, 2021 at 11:33 AM, Nicolas Hug <niourf@gmail.com> wrote:
To add to Guillaume's answer: the native multiclass support for forests/trees is described here: https://scikit-learn.org/stable/modules/tree.html#multi-output-problems <https://scikit-learn.org/stable/modules/tree.html#multi-output-problems> It's not a one-vs-rest strategy and can be summed up as:
Store n output values in leaves, instead of 1;
Use splitting criteria that compute the average reduction across all n outputs.
Nicolas
On 27/07/2021 10:22, Guillaume Lemaître wrote:
On 27 Jul 2021, at 11:08, Sole Galli via scikit-learn <scikit-learn@python.org> <mailto:scikit-learn@python.org> wrote:
Hello community,
Do I understand correctly that Random Forests are trained as a 1 vs rest when the target has more than 2 classes? Say the target takes values 0, 1 and 2, then the model would train 3 estimators 1 per class under the hood?. Each decision tree of the forest is natively supporting multi class.
The predict_proba output is an array with 3 columns, containing the probability of each class. If it is 1 vs rest. am I correct to assume that the sum of the probabilities for the 3 classes should not necessarily add up to 1? are they normalized? how is it done so that they do add up to 1? According to the above answer, the sum for each row of the array given by `predict_proba` will sum to 1. According to the documentation, the probabilities are computed as:
The predicted class probabilities of an input sample are computed as the mean predicted class probabilities of the trees in the forest. The class probability of a single tree is the fraction of samples of the same class in a leaf.
Thank you Sole
_______________________________________________ scikit-learn mailing list scikit-learn@python.org <mailto:scikit-learn@python.org> https://mail.python.org/mailman/listinfo/scikit-learn <https://mail.python.org/mailman/listinfo/scikit-learn>
scikit-learn mailing list scikit-learn@python.org <mailto:scikit-learn@python.org> https://mail.python.org/mailman/listinfo/scikit-learn <https://mail.python.org/mailman/listinfo/scikit-learn>
_______________________________________________ scikit-learn mailing list scikit-learn@python.org https://mail.python.org/mailman/listinfo/scikit-learn
2021年7月27日(火) 12:03 Guillaume Lemaître <g.lemaitre58@gmail.com>:
As far that I remember, `precision_recall_curve` and `roc_curve` do not support multi class. They are design to work only with binary classification.
Correct, the TPR-FPR curve (ROC) was originally intended for tuning a free parameter, in signal detection, and is a binary-type metric. For ML problems, it lets you tune/determine an estimator's output value threshold (e.g., a probability or a raw discriminant value such as in SVM) for arriving an optimized model that will be used to give a final, binary-discretized answer in new prediction tasks. Hope this helps, J.B.
Greetings! I am currently out of office, with limited access to emails, till August the 30th. Please contact support@giotto.ai for technical issue concerning Giotto Platform. Otherwise, I will reply to your email as soon as possible upon my return. With best regards, Matteo On 27 Jul 2021, at 12:42, Brown J.B. via scikit-learn <scikit-learn@python.org> wrote: 2021年7月27日(火) 12:03 Guillaume Lemaître <g.lemaitre58@gmail.com>: As far that I remember, `precision_recall_curve` and `roc_curve` do not support multi class. They are design to work only with binary classification. Correct, the TPR-FPR curve (ROC) was originally intended for tuning a free parameter, in signal detection, and is a binary-type metric. For ML problems, it lets you tune/determine an estimator's output value threshold (e.g., a probability or a raw discriminant value such as in SVM) for arriving an optimized model that will be used to give a final, binary-discretized answer in new prediction tasks. Hope this helps, J.B. _______________________________________________ scikit-learn mailing list scikit-learn@python.org https://mail.python.org/mailman/listinfo/scikit-learn
Greetings! I am currently out of office, with limited access to emails, till August the 30th. Please contact support@giotto.ai for technical issues concerning Giotto Platform. Otherwise, I will reply to your email as soon as possible upon my return. With best regards, Matteo On 27 Jul 2021, at 12:42, Brown J.B. via scikit-learn <scikit-learn@python.org> wrote: 2021年7月27日(火) 12:03 Guillaume Lemaître <g.lemaitre58@gmail.com>: As far that I remember, `precision_recall_curve` and `roc_curve` do not support multi class. They are design to work only with binary classification. Correct, the TPR-FPR curve (ROC) was originally intended for tuning a free parameter, in signal detection, and is a binary-type metric. For ML problems, it lets you tune/determine an estimator's output value threshold (e.g., a probability or a raw discriminant value such as in SVM) for arriving an optimized model that will be used to give a final, binary-discretized answer in new prediction tasks. Hope this helps, J.B. _______________________________________________ scikit-learn mailing list scikit-learn@python.org https://mail.python.org/mailman/listinfo/scikit-learn
Yellowbrick has multi label precision recall curves and multiclass roc/auc builtin: https://www.scikit-yb.org/en/latest/api/classifier/rocauc.html Sent from my iPad
On Jul 27, 2021, at 6:03 AM, Guillaume Lemaître <g.lemaitre58@gmail.com> wrote:
As far that I remember, `precision_recall_curve` and `roc_curve` do not support multi class. They are design to work only with binary classification. Then, we provide an example for precision-recall that shows one way to compute precision-recall curve via averaging: https://scikit-learn.org/stable/auto_examples/model_selection/plot_precision... -- Guillaume Lemaitre Scikit-learn @ Inria Foundation https://glemaitre.github.io/
On 27 Jul 2021, at 11:42, Sole Galli via scikit-learn <scikit-learn@python.org> wrote:
Thank you!
So when in the multiclass document says that for the algorithms that support intrinsically multiclass, which are listed here, when it says that they do not need to be wrapped by the OnevsRest, it means that there is no need, because they can indeed handle multi class, each one in their own way.
But, if I want to plot PR curves or ROC curves, then I do need to wrap them because those metrics are calculated as a 1 vs rest manner, and this is not how it is handled by the algos. Is my understanding correct?
Thank you!
‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐ On Tuesday, July 27th, 2021 at 11:33 AM, Nicolas Hug <niourf@gmail.com> wrote:
To add to Guillaume's answer: the native multiclass support for forests/trees is described here: https://scikit-learn.org/stable/modules/tree.html#multi-output-problems
It's not a one-vs-rest strategy and can be summed up as:
Store n output values in leaves, instead of 1;
Use splitting criteria that compute the average reduction across all n outputs.
Nicolas
On 27/07/2021 10:22, Guillaume Lemaître wrote:
On 27 Jul 2021, at 11:08, Sole Galli via scikit-learn <scikit-learn@python.org> wrote:
Hello community,
Do I understand correctly that Random Forests are trained as a 1 vs rest when the target has more than 2 classes? Say the target takes values 0, 1 and 2, then the model would train 3 estimators 1 per class under the hood?. Each decision tree of the forest is natively supporting multi class.
The predict_proba output is an array with 3 columns, containing the probability of each class. If it is 1 vs rest. am I correct to assume that the sum of the probabilities for the 3 classes should not necessarily add up to 1? are they normalized? how is it done so that they do add up to 1? According to the above answer, the sum for each row of the array given by `predict_proba` will sum to 1. According to the documentation, the probabilities are computed as:
The predicted class probabilities of an input sample are computed as the mean predicted class probabilities of the trees in the forest. The class probability of a single tree is the fraction of samples of the same class in a leaf.
Thank you Sole
_______________________________________________ scikit-learn mailing list scikit-learn@python.org https://mail.python.org/mailman/listinfo/scikit-learn
scikit-learn mailing list scikit-learn@python.org https://mail.python.org/mailman/listinfo/scikit-learn
_______________________________________________ scikit-learn mailing list scikit-learn@python.org https://mail.python.org/mailman/listinfo/scikit-learn
_______________________________________________ scikit-learn mailing list scikit-learn@python.org https://mail.python.org/mailman/listinfo/scikit-learn
participants (6)
-
Brown J.B. -
Francois Dion -
Guillaume Lemaître -
Matteo Caorsi -
Nicolas Hug -
Sole Galli