Name | cleanlab JSON |

Version | 1.0 JSON |

download | |

home_page | https://github.com/cgnorthcutt/cleanlab |

Summary | The standard package for machine learning with noisy labels and finding mislabeled data in Python. |

upload_time | 2021-04-18 19:42:44 |

maintainer | |

docs_url | None |

author | Curtis G. Northcutt |

requires_python | >=2.7,!=3.0.*,!=3.1.*,!=3.2.*,!=3.3.* |

license | GPLv3+ |

keywords | machine_learning denoising classification weak_supervision learning_with_noisy_labels unsupervised_learning |

VCS | |

bugtrack_url | |

requirements | No requirements were recorded. |

Travis-CI | No Travis. |

coveralls test coverage | No coveralls. |

.. figure:: https://raw.githubusercontent.com/cgnorthcutt/cleanlab/master/img/cleanlab_logo.png :target: https://github.com/cgnorthcutt/cleanlab/ :align: center :alt: cleanlab | ``cleanlab`` is python package for **machine learning with noisy labels**. ``cleanlab`` ``clean``\s ``lab``\els and supports **finding, quantifying, and learning** with label errors in datasets. View the: `cleanlab code documenation <https://cleanlab.readthedocs.io/>`__. ``cleanlab`` is powered by **confident learning**, published in this `paper <https://jair.org/index.php/jair/article/view/12125>`__ | `blog <https://l7.curtisnorthcutt.com/confident-learning>`__. **** * **News! (Apr 2021)** `cleanlab` published in the `confident learning paper (Northcutt, Jiang, & Chuang, 2021) <https://jair.org/index.php/jair/article/view/12125>`__ in the Journal of AI Research. * **News! (Mar 2021)** `cleanlab` supports ICLR workshop paper `(Northcutt, Athalye, & Mueller, 2021) <https://arxiv.org/abs/1911.00068>`__, by finding label errors across 10 common benchark datasets (ImageNet, CIFAR-10, CIFAR-100, Caltech-256, Quickdraw, MNIST, Amazon Reviews, IMDB, 20 News Groups, AudioSet). Along with the paper, the authors launched `labelerrors.com <https://labelerrors.com>`__ where you can view the label errors in these datasets. * **News! (Dec 2020)** `cleanlab` supports NeurIPS workshop paper `(Northcutt, Athalye, & Lin, 2020) <http://securedata.lol/camera_ready/28.pdf>`__. * **News! (Dec 2020)** `cleanlab` supports `PU learning <https://github.com/cgnorthcutt/cleanlab#pu-learning-with-cleanlab>`__. * **News! (Jan 2020)** `cleanlab` achieves state-of-the-art on CIFAR-10 for learning with noisy labels. Code to reproduce is here: `examples/cifar10 <https://github.com/cgnorthcutt/cleanlab/tree/master/examples/cifar10>`__. This is a great place for newcomers to see how to use cleanlab on real datasets. Data needed is available in the `confidentlearning-reproduce <https://github.com/cgnorthcutt/confidentlearning-reproduce>`__ repo, ``cleanlab`` v0.1.0 reproduces results in `the CL paper <https://arxiv.org/abs/1911.00068>`__. * **News! (Feb 2020)** `cleanlab` now natively supports Mac, Linux, and Windows. * **News! (Feb 2020)** `cleanlab` now supports `Co-Teaching <https://github.com/cgnorthcutt/cleanlab/blob/master/cleanlab/coteaching.py>`__ `(Han et al., 2018) <https://arxiv.org/abs/1804.06872>`__. |pypi| |os| |py_versions| |build_status| |coverage| .. |pypi| image:: https://img.shields.io/pypi/v/cleanlab.svg :target: https://pypi.org/pypi/cleanlab/ .. |os| image:: https://img.shields.io/badge/platform-windows%20%7C%20macos%20%7C%20linux-lightgrey :target: https://pypi.org/pypi/cleanlab/ .. |py_versions| image:: https://img.shields.io/pypi/pyversions/cleanlab.svg :target: https://pypi.org/pypi/cleanlab/ .. |build_status| image:: https://travis-ci.com/cgnorthcutt/cleanlab.svg?branch=master :target: https://travis-ci.com/cgnorthcutt/cleanlab .. |coverage| image:: https://codecov.io/gh/cgnorthcutt/cleanlab/branch/master/graph/badge.svg :target: https://codecov.io/gh/cgnorthcutt/cleanlab ``cleanlab`` **documentation** is available in `this blog post <https://l7.curtisnorthcutt.com/cleanlab-python-package>`__. Past release notes and **future features planned** is available `here <https://github.com/cgnorthcutt/cleanlab/blob/master/cleanlab/version.py>`__. So fresh, so ``cleanlab`` ========================= ``cleanlab`` finds and cleans label errors in any dataset using `state-of-the-art algorithms <https://arxiv.org/abs/1911.00068>`__ to find label errors, characterize noise, and learn in spite of it. ``cleanlab`` is fast: its built on optimized algorithms and parallelized across CPU threads automatically. ``cleanlab`` is powered by `provable guarantees <https://arxiv.org/abs/1911.00068>`__ of exact noise estimation and label error finding in realistic cases when model output probabilities are erroneous. ``cleanlab`` supports multi-label, multiclass, sparse matrices, etc. By default, ``cleanlab`` requires no hyper-parameters. ``cleanlab`` implements the family of theory and algorithms called `confident learning <https://arxiv.org/abs/1911.00068>`__ with provable guarantees of exact noise estimation and label error finding (even when model output probabilities are noisy/imperfect). **How does confident learning work?** See: `TUTORIAL: confident learning with just numpy and for-loops <https://github.com/cgnorthcutt/cleanlab/blob/master/examples/simplifying_confident_learning_tutorial.ipynb>`__. ``cleanlab`` supports most weak supervision tasks: multi-label, multiclass, sparse matrices, etc. ``cleanlab`` is: 1. backed-by-theory - Provable perfect label error finding in realistic conditions. 2. fast - Non-iterative, parallelized algorithms (e.g. < 1 second to find label errors in ImageNet) 3. general - Works with any ML or deep learning framework: PyTorch, Tensorflow, MxNet, Caffe2, scikit-learn, etc. 4. unique - The only package for weak supervion with any dataset / classifier. Find label errors with PyTorch, Tensorflow, MXNet, etc. in 1 line of code. ========================================================================== .. code:: python # Compute psx (n x m matrix of predicted probabilities) on your own, with any classifier. # Here is an example that shows in detail how to compute psx on CIFAR-10: # https://github.com/cgnorthcutt/cleanlab/tree/master/examples/cifar10 # Be sure you compute probs in a holdout/out-of-sample manner (e.g. cross-validation) # Now getting label errors is trivial with cleanlab... its one line of code. # Label errors are ordered by likelihood of being an error. First index is most likely error. from cleanlab.pruning import get_noise_indices ordered_label_errors = get_noise_indices( s=numpy_array_of_noisy_labels, psx=numpy_array_of_predicted_probabilities, sorted_index_method='normalized_margin', # Orders label errors ) Pre-computed out-of-sample predicted probabilities for CIFAR-10 train set are available here: [`[LINK] <https://github.com/cgnorthcutt/cleanlab/tree/master/examples/cifar10>`__]. Learning with noisy labels in 3 lines of code! ============================================== .. code:: python from cleanlab.classification import LearningWithNoisyLabels from sklearn.linear_model import LogisticRegression # Wrap around any classifier. Yup, you can use sklearn/pyTorch/Tensorflow/FastText/etc. lnl = LearningWithNoisyLabels(clf=LogisticRegression()) lnl.fit(X=X_train_data, s=train_noisy_labels) # Estimate the predictions you would have gotten by training with *no* label errors. predicted_test_labels = lnl.predict(X_test) Check out these `examples <https://github.com/cgnorthcutt/cleanlab/tree/master/examples>`__ and `tests <https://github.com/cgnorthcutt/cleanlab/tree/master/tests>`__ (includes how to use pyTorch, FastText, etc.). Installation ============ Python 2.7, 3.4, 3.5, 3.6, and 3.7 are supported. Linux, macOS, and Windows are supported. Stable release: .. code-block:: bash $ pip install cleanlab Developer (unstable) release: .. code-block:: bash $ pip install git+https://github.com/cgnorthcutt/cleanlab.git To install the codebase (enabling you to make modifications): .. code-block:: bash $ conda update pip # if you use conda $ git clone https://github.com/cgnorthcutt/cleanlab.git $ cd cleanlab $ pip install -e . Citations and Related Publications ================================== If you use this package, please cite the `confident learning paper <https://arxiv.org/abs/1911.00068>`__ (published April 2021, in the `Journal of AI Research <https://jair.org/index.php/jair/article/view/12125>`__): :: @article{northcutt2021confidentlearning, title={Confident Learning: Estimating Uncertainty in Dataset Labels}, author={Curtis G. Northcutt and Lu Jiang and Isaac L. Chuang}, journal={Journal of Artificial Intelligence Research (JAIR)}, volume={70}, pages={1373--1411}, year={2021} } If you use this package for binary classification, please also cite the `rankpruning paper <https://arxiv.org/abs/1705.01936>`__ (published August 2017, in `Uncertainty in AI <http://auai.org/uai2017/proceedings/papers/35.pdf>`__): :: @inproceedings{northcutt2017rankpruning, author={Northcutt, Curtis G. and Wu, Tailin and Chuang, Isaac L.}, title={Learning with Confident Examples: Rank Pruning for Robust Classification with Noisy Labels}, booktitle = {Proceedings of the Thirty-Third Conference on Uncertainty in Artificial Intelligence}, series = {UAI'17}, year = {2017}, location = {Sydney, Australia}, numpages = {10}, url = {http://auai.org/uai2017/proceedings/papers/35.pdf}, publisher = {AUAI Press}, } Reproducing Results in `confident learning paper <https://arxiv.org/abs/1911.00068>`__ ======================================================================================= See `cleanlab/examples/cifar10 <https://github.com/cgnorthcutt/cleanlab/tree/master/examples/cifar10>`__ and `cleanlab/examples/imagenet <https://github.com/cgnorthcutt/cleanlab/tree/master/examples/imagenet>`__. You'll need to ``git clone`` `confidentlearning-reproduce <https://github.com/cgnorthcutt/confidentlearning-reproduce>`__ which contains the data and files needed to reproduce the CIFAR-10 results. ``cleanlab``: State of the Art Learning with Noisy Labels in CIFAR ------------------------------------------------------------------ A [`step-by-step guide <https://github.com/cgnorthcutt/cleanlab/tree/master/examples/cifar10>`__] to reproduce these results is available [`here <https://github.com/cgnorthcutt/cleanlab/tree/master/examples/cifar10>`__]. This guide is also helpful as a tutorial to use cleanlab on any large-scale dataset. .. figure:: https://raw.githubusercontent.com/cgnorthcutt/cleanlab/master/img/cifar10_benchmarks.png :align: center :alt: Image depicting CIFAR10 benchmarks Comparison of confident learning (CL) and `cleanlab` versus seven recent methods for learning with noisy labels in CIFAR-10. Highlighted cells show CL robustness to sparsity. The five CL methods estimate label errors, remove them, then train on the cleaned data using `Co-Teaching <https://github.com/cgnorthcutt/cleanlab/blob/master/cleanlab/coteaching.py>`__. Observe how cleanlab (CL methods) are robust to large sparsity in label noise whereas prior art tends to reduce in performance for increased sparsity, as shown by the red highlighted regions. This is important because real-world label noise is often sparse, e.g. a tiger is likely to be mislabeled as a lion, but not as most other classes like airplane, bathtub, and microwave. ``cleanlab``: Find Label Errors in ImageNet ------------------------------------------- Use ``cleanlab`` to identify ~100,000 label errors in the 2012 ImageNet training dataset. .. figure:: https://raw.githubusercontent.com/cgnorthcutt/cleanlab/master/img/imagenet_train_label_errors_32.jpg :align: center :alt: Image depicting label errors in ImageNet train set Top label issues in the 2012 ILSVRC ImageNet train set identified using ``cleanlab``. Label Errors are boxed in red. Ontological issues in green. Multi-label images in blue. ``cleanlab``: Find Label Errors in MNIST ---------------------------------------- Use ``cleanlab`` to identify ~50 label errors in the MNIST dataset. .. figure:: https://raw.githubusercontent.com/cgnorthcutt/cleanlab/master/img/mnist_training_label_errors24_prune_by_noise_rate.png :align: center :alt: Image depicting label errors in MNIST train set Label errors of the original MNIST **train** dataset identified algorithmically using cleanlab. Depicts the 24 least confident labels, ordered left-right, top-down by increasing self-confidence (probability of belonging to the given label), denoted conf in teal. The label with the largest predicted probability is in green. Overt errors are in red. ``cleanlab`` Generality: View performance across 4 distributions and 9 classifiers. ----------------------------------------------------------------------------------- Use ``cleanlab`` to learn with noisy labels regardless of dataset distribution or classifier. .. figure:: https://raw.githubusercontent.com/cgnorthcutt/cleanlab/master/img/demo_cleanlab_across_datasets_and_classifiers.png :align: center :alt: Image depicting generality of cleanlab across datasets and classifiers Each sub-figure in the figure above depicts the decision boundary learned using ``cleanlab.classification.LearningWithNoisyLabels`` in the presence of extreme (\~35%) label errors. Label errors are circled in green. Label noise is class-conditional (not simply uniformly random). Columns are organized by the classifier used, except the left-most column which depicts the ground-truth dataset distribution. Rows are organized by dataset used. The code to reproduce this figure is available `here <https://github.com/cgnorthcutt/cleanlab/blob/master/examples/classifier_comparison.ipynb>`__. Each figure depicts accuracy scores on a test set as decimal values: 1. LEFT (in black): The classifier test accuracy trained with perfect labels (no label errors). 2. MIDDLE (in blue): The classifier test accuracy trained with noisy labels using ``cleanlab``. 3. RIGHT (in white): The baseline classifier test accuracy trained with noisy labels. As an example, this is the noise matrix (noisy channel) *P(s \| y)* characterizing the label noise for the first dataset row in the figure. *s* represents the observed noisy labels and *y* represents the latent, true labels. The trace of this matrix is 2.6. A trace of 4 implies no label noise. A cell in this matrix is read like, "A random 38% of '3' labels were flipped to '2' labels." ====== ==== ==== ==== ==== p(s|y) y=0 y=1 y=2 y=3 ====== ==== ==== ==== ==== s=0 0.55 0.01 0.07 0.06 s=1 0.22 0.87 0.24 0.02 s=2 0.12 0.04 0.64 0.38 s=3 0.11 0.08 0.05 0.54 ====== ==== ==== ==== ==== Get started with easy, quick examples. ====================================== New to **cleanlab**? Start with: 1. `Visualizing confident learning <https://github.com/cgnorthcutt/cleanlab/blob/master/examples/visualizing_confident_learning.ipynb>`__ 2. `A simple example of learning with noisy labels on the multiclass Iris dataset <https://github.com/cgnorthcutt/cleanlab/blob/master/examples/iris_simple_example.ipynb>`__. These examples show how easy it is to characterize label noise in datasets, learn with noisy labels, identify label errors, estimate latent priors and noisy channels, and more. .. .. <!--- ![Image depicting label errors in MNIST test set.](https://raw.githubusercontent.com/cgnorthcutt/cleanlab/master/img/mnist_test_label_errors8.png) Selected label errors in the MNIST **test** dataset ordered by increasing self-confidence (in teal). ## Automatically identify ~5k (of 50k) validation set label errors in ImageNet. [[link]](examples/finding_ImageNet_label_errors). ![Image depicting label errors in ImageNet validation set.](https://raw.githubusercontent.com/cgnorthcutt/cleanlab/master/img/imagenet_validation_label_errors_96_prune_by_noise_rate.jpg) Label errors in the 2012 ImageNet validation dataset identified automatically with cleanlab using a pre-trained resnet18. Displayed are the 96 least confident labels. We see that ImageNet contains numerous multi-label images, although it is used widely by the machine learning and vision communities as a single-label benchmark dataset. ---> Use ``cleanlab`` with any model (Tensorflow, caffe2, PyTorch, etc.) ------------------------------------------------------------------- All of the features of the ``cleanlab`` package work with **any model**. Yes, any model. Feel free to use PyTorch, Tensorflow, caffe2, scikit-learn, mxnet, etc. If you use a scikit-learn classifier, all ``cleanlab`` methods will work out-of-the-box. It’s also easy to use your favorite model from a non-scikit-learn package, just wrap your model into a Python class that inherits the ``sklearn.base.BaseEstimator``: .. code:: python from sklearn.base import BaseEstimator class YourFavoriteModel(BaseEstimator): # Inherits sklearn base classifier def __init__(self, ): pass def fit(self, X, y, sample_weight=None): pass def predict(self, X): pass def predict_proba(self, X): pass def score(self, X, y, sample_weight=None): pass # Now you can use your model with `cleanlab`. Here's one example: from cleanlab.classification import LearningWithNoisyLabels lnl = LearningWithNoisyLabels(clf=YourFavoriteModel()) lnl.fit(train_data, train_labels_with_errors) Want to see a working example? `Here’s a compliant PyTorch MNIST CNN class <https://github.com/cgnorthcutt/cleanlab/blob/master/cleanlab/models/mnist_pytorch.py#L28>`__ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ As you can see `here <https://github.com/cgnorthcutt/cleanlab/blob/master/cleanlab/models/mnist_pytorch.py#L28>`__, technically you don’t actually need to inherit from ``sklearn.base.BaseEstimator``, as you can just create a class that defines .fit(), .predict(), and .predict_proba(), but inheriting makes downstream scikit-learn applications like hyper-parameter optimization work seamlessly. For example, the `LearningWithNoisyLabels() model <https://github.com/cgnorthcutt/cleanlab/blob/master/cleanlab/classification.py#L48>`__ is fully compliant. Note, some libraries exists to do this for you. For pyTorch, check out the ``skorch`` Python library which will wrap your ``pytorch`` model into a ``scikit-learn`` compliant model. Documentation by Example ======================== ``cleanlab`` Core Package Components ------------------------------------ 1. **cleanlab/classification.py** - The LearningWithNoisyLabels() class for learning with noisy labels. 2. **cleanlab/latent_algebra.py** - Equalities when noise information is known. 3. **cleanlab/latent_estimation.py** - Estimates and fully characterizes all variants of label noise. 4. **cleanlab/noise_generation.py** - Generate mathematically valid synthetic noise matrices. 5. **cleanlab/polyplex.py** - Characterizes joint distribution of label noise EXACTLY from noisy channel. 6. **cleanlab/pruning.py** - Finds the indices of the examples with label errors in a dataset. Many of these methods have default parameters that won’t be covered here. Check out the method docstrings for full documentation. Methods to Standardize Research with Noisy Labels ------------------------------------------------- ``cleanlab`` supports a number of functions to generate noise for benchmarking and standardization in research. This next example shows how to generate valid, class-conditional, unformly random noisy channel matrices: .. code:: python # Generate a valid (necessary conditions for learnability are met) noise matrix for any trace > 1 from cleanlab.noise_generation import generate_noise_matrix_from_trace noise_matrix=generate_noise_matrix_from_trace( K=number_of_classes, trace=float_value_greater_than_1_and_leq_K, py=prior_of_y_actual_labels_which_is_just_an_array_of_length_K, frac_zero_noise_rates=float_from_0_to_1_controlling_sparsity, ) # Check if a noise matrix is valid (necessary conditions for learnability are met) from cleanlab.noise_generation import noise_matrix_is_valid is_valid=noise_matrix_is_valid(noise_matrix, prior_of_y_which_is_just_an_array_of_length_K) For a given noise matrix, this example shows how to generate noisy labels. Methods can be seeded for reproducibility. .. code:: python # Generate noisy labels using the noise_marix. Guarantees exact amount of noise in labels. from cleanlab.noise_generation import generate_noisy_labels s_noisy_labels = generate_noisy_labels(y_hidden_actual_labels, noise_matrix) # This package is a full of other useful methods for learning with noisy labels. # The tutorial stops here, but you don't have to. Inspect method docstrings for full docs. Estimate the confident joint, the latent noisy channel matrix, *P(s \| y)* and inverse, *P(y \| s)*, the latent prior of the unobserved, actual true labels, *p(y)*, and the predicted probabilities. ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ *s* denotes a random variable that represents the observed, noisy label and *y* denotes a random variable representing the hidden, actual labels. Both *s* and *y* take any of the m classes as values. The ``cleanlab`` package supports different levels of granularity for computation depending on the needs of the user. Because of this, we support multiple alternatives, all no more than a few lines, to estimate these latent distribution arrays, enabling the user to reduce computation time by only computing what they need to compute, as seen in the examples below. Throughout these examples, you’ll see a variable called *confident_joint*. The confident joint is an m x m matrix (m is the number of classes) that counts, for every observed, noisy class, the number of examples that confidently belong to every latent, hidden class. It counts the number of examples that we are confident are labeled correctly or incorrectly for every pair of obseved and unobserved classes. The confident joint is an unnormalized estimate of the complete-information latent joint distribution, *Ps,y*. Most of the methods in the **cleanlab** package start by first estimating the *confident_joint*. You can learn more about this in the `confident learning paper <https://arxiv.org/abs/1911.00068>`__. Option 1: Compute the confident joint and predicted probs first. Stop if that’s all you need. ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ .. code:: python from cleanlab.latent_estimation import estimate_latent from cleanlab.latent_estimation import estimate_confident_joint_and_cv_pred_proba # Compute the confident joint and the n x m predicted probabilities matrix (psx), # for n examples, m classes. Stop here if all you need is the confident joint. confident_joint, psx = estimate_confident_joint_and_cv_pred_proba( X=X_train, s=train_labels_with_errors, clf=logreg(), # default, you can use any classifier ) # Estimate latent distributions: p(y) as est_py, P(s|y) as est_nm, and P(y|s) as est_inv est_py, est_nm, est_inv = estimate_latent(confident_joint, s=train_labels_with_errors) Option 2: Estimate the latent distribution matrices in a single line of code. ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ .. code:: python from cleanlab.latent_estimation import estimate_py_noise_matrices_and_cv_pred_proba est_py, est_nm, est_inv, confident_joint, psx = estimate_py_noise_matrices_and_cv_pred_proba( X=X_train, s=train_labels_with_errors, ) Option 3: Skip computing the predicted probabilities if you already have them. ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ .. code:: python # Already have psx? (n x m matrix of predicted probabilities) # For example, you might get them from a pre-trained model (like resnet on ImageNet) # With the cleanlab package, you estimate directly with psx. from cleanlab.latent_estimation import estimate_py_and_noise_matrices_from_probabilities est_py, est_nm, est_inv, confident_joint = estimate_py_and_noise_matrices_from_probabilities( s=train_labels_with_errors, psx=psx, ) Completely characterize label noise in a dataset: ------------------------------------------------- The joint probability distribution of noisy and true labels, *P(s,y)*, completely characterizes label noise with a class-conditional *m x m* matrix. .. code:: python from cleanlab.latent_estimation import estimate_joint joint = estimate_joint( s=noisy_labels, psx=probabilities, confident_joint=None, # Provide if you have it already ) PU learning with cleanlab: -------------------------- PU learning is a special case when one of your classes has no error. P stands for the positive class and **is assumed to have zero label errors** and U stands for unlabeled data, but in practice, we just assume the U class is a noisy negative class that contains some positive examples. Thus, the goal of PU learning is to (1) estimate the proportion of positives in the negative class (see `fraction_noise_in_unlabeled_class` in the last example), (2) find the errors (see last example), and (3) train on clean data (see first example below). `cleanlab` does all three, taking into account that there is no label errors in whichever class you specify. There are two ways to use `cleanlab` for PU learning. We'll look at each here. Method 1. If you are using the cleanlab classifier `LearningWithNoisyLabels()`, and your dataset has exactly two classes (positive = 1, and negative = 0), PU learning is supported directly in `cleanlab`. You can perform PU learning like this: .. code:: python from cleanlab.classification import LearningWithNoisyLabels from sklearn.linear_model import LogisticRegression # Wrap around any classifier. Yup, you can use sklearn/pyTorch/Tensorflow/FastText/etc. pu_class = 0 # Should be 0 or 1. Label of class with NO ERRORS. (e.g., P class in PU) lnl = LearningWithNoisyLabels(clf=LogisticRegression(), pulearning=pu_class) lnl.fit(X=X_train_data, s=train_noisy_labels) # Estimate the predictions you would have gotten by training with *no* label errors. predicted_test_labels = lnl.predict(X_test) Method 2. However, you might be using a more complicated classifier that doesn't work well with LearningWithNoisyLabels (see this example for CIFAR-10). Or you might have 3 or more classes. Here's how to use cleanlab for PU learning in this situation. To let cleanlab know which class has no error (in standard PU learning, this is the P class), you need to set the threshold for that class to 1 (1 means the probability that the labels of that class are correct is 1, i.e. that class has no error). Here's the code: .. code:: python import numpy as np # K is the number of classes in your dataset # psx are the cross-validated predicted probabilities. # s is the array/list/iterable of noisy labels # pu_class is a 0-based integer for the class that has no label errors. thresholds = np.asarray([np.mean(psx[:, k][s == k]) for k in range(K)]) thresholds[pu_class] = 1.0 Now you can use cleanlab however you were before. Just be sure to pass in this thresholds parameter wherever it applies. For example: .. code:: python # Uncertainty quantification (characterize the label noise # by estimating the joint distribution of noisy and true labels) cj = compute_confident_joint(s, psx, thresholds=thresholds, ) # Now the noise (cj) has been estimated taking into account that some class(es) have no error. # We can use cj to find label errors like this: indices_of_label_errors = get_noise_indices(s, psx, confident_joint=cj, ) # In addition to label errors, we can find the fraction of noise in the unlabeled class. # First we need the inv_noise_matrix which contains P(y|s) (proportion of mislabeling). _, _, inv_noise_matrix = estimate_latent(confident_joint=cj, s=s, ) # Because inv_noise_matrix contains P(y|s), p (y = anything | s = pu_class) should be 0 # because the prob(true label is something else | example is in pu_class) is 0. # What's more interesting is p(y = anything | s is not put_class), or in the binary case # this translates to p(y = pu_class | s = 1 - pu_class) because pu_class is 0 or 1. # So, to find the fraction_noise_in_unlabeled_class, for binary, you just compute: fraction_noise_in_unlabeled_class = inv_noise_matrix[pu_class][1 - pu_class] Now that you have `indices_of_label_errors`, you can remove those label errors and train on clean data (or only remove some of the label errors and iteratively use confident learning / cleanlab to improve results) The Polyplex ------------ The key to learning in the presence of label errors is estimating the joint distribution between the actual, hidden labels ‘*y*’ and the observed, noisy labels ‘*s*’. Using ``cleanlab`` and the theory of confident learning, we can completely characterize the trace of the latent joint distribution, *trace(P(s,y))*, given *p(y)*, for any fraction of label errors, i.e. for any trace of the noisy channel, *trace(P(s|y))*. You can check out how to do this yourself here: 1. `Drawing Polyplices <https://github.com/cgnorthcutt/cleanlab/blob/master/examples/drawing_polyplices.ipynb>`__ 2. `Computing Polyplices <https://github.com/cgnorthcutt/cleanlab/blob/master/cleanlab/polyplex.py>`__ Join our community ------------------ Have ideas for the future of cleanlab? How are you using cleanlab? `Join the discussion <https://github.com/cgnorthcutt/cleanlab/discussions>`__. Have code improvements for cleanlab? `Submit a code pull request <https://github.com/cgnorthcutt/cleanlab/issues/new>`__. Do you have an issue with cleanlab? `Submit an issue <https://github.com/cgnorthcutt/cleanlab/issues/new>`__. License ------- Copyright (c) 2017-2050 Curtis G. Northcutt cleanlab is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. cleanlab is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See `GNU General Public LICENSE <https://github.com/cgnorthcutt/cleanlab/blob/master/LICENSE>`__ for details. THIS LICENSE APPLIES TO THIS VERSION AND ALL PREVIOUS VERSIONS OF cleanlab.

{ "_id": null, "home_page": "https://github.com/cgnorthcutt/cleanlab", "name": "cleanlab", "maintainer": "", "docs_url": null, "requires_python": ">=2.7,!=3.0.*,!=3.1.*,!=3.2.*,!=3.3.*", "maintainer_email": "", "keywords": "machine_learning denoising classification weak_supervision learning_with_noisy_labels unsupervised_learning", "author": "Curtis G. Northcutt", "author_email": "cgn@csail.mit.edu", "download_url": "https://files.pythonhosted.org/packages/82/56/9f4c5bc333bff69b56a1bcbbee73e751b0fd997b7960e089fecbdaa79726/cleanlab-1.0.tar.gz", "platform": "", "description": ".. figure:: https://raw.githubusercontent.com/cgnorthcutt/cleanlab/master/img/cleanlab_logo.png\n :target: https://github.com/cgnorthcutt/cleanlab/\n :align: center\n :alt: cleanlab \n\n| \n\n``cleanlab`` is python package for **machine learning with noisy labels**. ``cleanlab`` ``clean``\\s ``lab``\\els and supports **finding, quantifying, and learning** with label errors in datasets. View the: `cleanlab code documenation <https://cleanlab.readthedocs.io/>`__. \n\n``cleanlab`` is powered by **confident learning**, published in this `paper <https://jair.org/index.php/jair/article/view/12125>`__ | `blog <https://l7.curtisnorthcutt.com/confident-learning>`__. \n\n****\n\n* **News! (Apr 2021)** `cleanlab` published in the `confident learning paper (Northcutt, Jiang, & Chuang, 2021) <https://jair.org/index.php/jair/article/view/12125>`__ in the Journal of AI Research.\n* **News! (Mar 2021)** `cleanlab` supports ICLR workshop paper `(Northcutt, Athalye, & Mueller, 2021) <https://arxiv.org/abs/1911.00068>`__, by finding label errors across 10 common benchark datasets (ImageNet, CIFAR-10, CIFAR-100, Caltech-256, Quickdraw, MNIST, Amazon Reviews, IMDB, 20 News Groups, AudioSet). Along with the paper, the authors launched `labelerrors.com <https://labelerrors.com>`__ where you can view the label errors in these datasets.\n* **News! (Dec 2020)** `cleanlab` supports NeurIPS workshop paper `(Northcutt, Athalye, & Lin, 2020) <http://securedata.lol/camera_ready/28.pdf>`__.\n* **News! (Dec 2020)** `cleanlab` supports `PU learning <https://github.com/cgnorthcutt/cleanlab#pu-learning-with-cleanlab>`__.\n* **News! (Jan 2020)** `cleanlab` achieves state-of-the-art on CIFAR-10 for learning with noisy labels. Code to reproduce is here: `examples/cifar10 <https://github.com/cgnorthcutt/cleanlab/tree/master/examples/cifar10>`__. This is a great place for newcomers to see how to use cleanlab on real datasets. Data needed is available in the `confidentlearning-reproduce <https://github.com/cgnorthcutt/confidentlearning-reproduce>`__ repo, ``cleanlab`` v0.1.0 reproduces results in `the CL paper <https://arxiv.org/abs/1911.00068>`__.\n* **News! (Feb 2020)** `cleanlab` now natively supports Mac, Linux, and Windows.\n* **News! (Feb 2020)** `cleanlab` now supports `Co-Teaching <https://github.com/cgnorthcutt/cleanlab/blob/master/cleanlab/coteaching.py>`__ `(Han et al., 2018) <https://arxiv.org/abs/1804.06872>`__.\n\n\n|pypi| |os| |py_versions| |build_status| |coverage|\n\n.. |pypi| image:: https://img.shields.io/pypi/v/cleanlab.svg\n :target: https://pypi.org/pypi/cleanlab/\n.. |os| image:: https://img.shields.io/badge/platform-windows%20%7C%20macos%20%7C%20linux-lightgrey\n :target: https://pypi.org/pypi/cleanlab/\n.. |py_versions| image:: https://img.shields.io/pypi/pyversions/cleanlab.svg\n :target: https://pypi.org/pypi/cleanlab/\n.. |build_status| image:: https://travis-ci.com/cgnorthcutt/cleanlab.svg?branch=master\n :target: https://travis-ci.com/cgnorthcutt/cleanlab\n.. |coverage| image:: https://codecov.io/gh/cgnorthcutt/cleanlab/branch/master/graph/badge.svg\n :target: https://codecov.io/gh/cgnorthcutt/cleanlab\n\n``cleanlab`` **documentation** is available in `this blog post <https://l7.curtisnorthcutt.com/cleanlab-python-package>`__.\n\nPast release notes and **future features planned** is available `here <https://github.com/cgnorthcutt/cleanlab/blob/master/cleanlab/version.py>`__.\n\nSo fresh, so ``cleanlab`` \n=========================\n\n``cleanlab`` finds and cleans label errors in any dataset using `state-of-the-art algorithms <https://arxiv.org/abs/1911.00068>`__ to find label errors, characterize noise, and learn in spite of it. ``cleanlab`` is fast: its built on optimized algorithms and parallelized across CPU threads automatically. ``cleanlab`` is powered by `provable guarantees <https://arxiv.org/abs/1911.00068>`__ of exact noise estimation and label error finding in realistic cases when model output probabilities are erroneous. ``cleanlab`` supports multi-label, multiclass, sparse matrices, etc. By default, ``cleanlab`` requires no hyper-parameters.\n\n``cleanlab`` implements the family of theory and algorithms called `confident learning <https://arxiv.org/abs/1911.00068>`__ with provable guarantees of exact noise estimation and label error finding (even when model output probabilities are noisy/imperfect). \n\n**How does confident learning work?** See: `TUTORIAL: confident learning with just numpy and for-loops <https://github.com/cgnorthcutt/cleanlab/blob/master/examples/simplifying_confident_learning_tutorial.ipynb>`__.\n\n``cleanlab`` supports most weak supervision tasks: multi-label, multiclass, sparse matrices, etc. \n\n``cleanlab`` is:\n\n1. backed-by-theory - Provable perfect label error finding in realistic conditions.\n2. fast - Non-iterative, parallelized algorithms (e.g. < 1 second to find label errors in ImageNet)\n3. general - Works with any ML or deep learning framework: PyTorch, Tensorflow, MxNet, Caffe2, scikit-learn, etc.\n4. unique - The only package for weak supervion with any dataset / classifier.\n\n\nFind label errors with PyTorch, Tensorflow, MXNet, etc. in 1 line of code.\n==========================================================================\n\n.. code:: python\n\n # Compute psx (n x m matrix of predicted probabilities) on your own, with any classifier.\n # Here is an example that shows in detail how to compute psx on CIFAR-10:\n # https://github.com/cgnorthcutt/cleanlab/tree/master/examples/cifar10\n # Be sure you compute probs in a holdout/out-of-sample manner (e.g. cross-validation)\n # Now getting label errors is trivial with cleanlab... its one line of code.\n # Label errors are ordered by likelihood of being an error. First index is most likely error.\n from cleanlab.pruning import get_noise_indices\n\n ordered_label_errors = get_noise_indices(\n s=numpy_array_of_noisy_labels,\n psx=numpy_array_of_predicted_probabilities,\n sorted_index_method='normalized_margin', # Orders label errors\n )\n\nPre-computed out-of-sample predicted probabilities for CIFAR-10 train set are available here: [`[LINK] <https://github.com/cgnorthcutt/cleanlab/tree/master/examples/cifar10>`__].\n\nLearning with noisy labels in 3 lines of code!\n==============================================\n\n.. code:: python\n\n from cleanlab.classification import LearningWithNoisyLabels\n from sklearn.linear_model import LogisticRegression\n\n # Wrap around any classifier. Yup, you can use sklearn/pyTorch/Tensorflow/FastText/etc.\n lnl = LearningWithNoisyLabels(clf=LogisticRegression()) \n lnl.fit(X=X_train_data, s=train_noisy_labels) \n # Estimate the predictions you would have gotten by training with *no* label errors.\n predicted_test_labels = lnl.predict(X_test)\n\n\nCheck out these `examples <https://github.com/cgnorthcutt/cleanlab/tree/master/examples>`__ and `tests <https://github.com/cgnorthcutt/cleanlab/tree/master/tests>`__ (includes how to use pyTorch, FastText, etc.).\n\n\nInstallation\n============\n\nPython 2.7, 3.4, 3.5, 3.6, and 3.7 are supported. Linux, macOS, and Windows are supported.\n\nStable release:\n\n.. code-block:: bash\n\n $ pip install cleanlab\n\nDeveloper (unstable) release:\n\n.. code-block:: bash\n\n $ pip install git+https://github.com/cgnorthcutt/cleanlab.git\n\nTo install the codebase (enabling you to make modifications):\n\n.. code-block:: bash\n\n $ conda update pip # if you use conda\n $ git clone https://github.com/cgnorthcutt/cleanlab.git\n $ cd cleanlab\n $ pip install -e .\n\n\nCitations and Related Publications\n==================================\n\nIf you use this package, please cite the `confident learning paper <https://arxiv.org/abs/1911.00068>`__ (published April 2021, in the `Journal of AI Research <https://jair.org/index.php/jair/article/view/12125>`__):\n\n::\n\n @article{northcutt2021confidentlearning,\n title={Confident Learning: Estimating Uncertainty in Dataset Labels},\n author={Curtis G. Northcutt and Lu Jiang and Isaac L. Chuang},\n journal={Journal of Artificial Intelligence Research (JAIR)},\n volume={70},\n pages={1373--1411},\n year={2021}\n }\n\n\n\nIf you use this package for binary classification, please also cite the `rankpruning paper <https://arxiv.org/abs/1705.01936>`__ (published August 2017, in `Uncertainty in AI <http://auai.org/uai2017/proceedings/papers/35.pdf>`__):\n\n::\n\n @inproceedings{northcutt2017rankpruning,\n author={Northcutt, Curtis G. and Wu, Tailin and Chuang, Isaac L.},\n title={Learning with Confident Examples: Rank Pruning for Robust Classification with Noisy Labels},\n booktitle = {Proceedings of the Thirty-Third Conference on Uncertainty in Artificial Intelligence},\n series = {UAI'17},\n year = {2017},\n location = {Sydney, Australia},\n numpages = {10},\n url = {http://auai.org/uai2017/proceedings/papers/35.pdf},\n publisher = {AUAI Press},\n } \n\nReproducing Results in `confident learning paper <https://arxiv.org/abs/1911.00068>`__ \n=======================================================================================\n\nSee `cleanlab/examples/cifar10 <https://github.com/cgnorthcutt/cleanlab/tree/master/examples/cifar10>`__ and `cleanlab/examples/imagenet <https://github.com/cgnorthcutt/cleanlab/tree/master/examples/imagenet>`__. You'll need to ``git clone`` `confidentlearning-reproduce <https://github.com/cgnorthcutt/confidentlearning-reproduce>`__ which contains the data and files needed to reproduce the CIFAR-10 results.\n\n\n``cleanlab``: State of the Art Learning with Noisy Labels in CIFAR\n------------------------------------------------------------------\n\n\nA [`step-by-step guide <https://github.com/cgnorthcutt/cleanlab/tree/master/examples/cifar10>`__] to reproduce these results is available [`here <https://github.com/cgnorthcutt/cleanlab/tree/master/examples/cifar10>`__]. This guide is also helpful as a tutorial to use cleanlab on any large-scale dataset.\n\n.. figure:: https://raw.githubusercontent.com/cgnorthcutt/cleanlab/master/img/cifar10_benchmarks.png\n :align: center\n :alt: Image depicting CIFAR10 benchmarks \n\nComparison of confident learning (CL) and `cleanlab` versus seven recent methods for learning with noisy labels in CIFAR-10. Highlighted cells show CL robustness to sparsity. The five CL methods estimate label errors, remove them, then train on the cleaned data using `Co-Teaching <https://github.com/cgnorthcutt/cleanlab/blob/master/cleanlab/coteaching.py>`__.\n\nObserve how cleanlab (CL methods) are robust to large sparsity in label noise whereas prior art tends to reduce in performance for increased sparsity, as shown by the red highlighted regions. This is important because real-world label noise is often sparse, e.g. a tiger is likely to be mislabeled as a lion, but not as most other classes like airplane, bathtub, and microwave.\n\n``cleanlab``: Find Label Errors in ImageNet\n-------------------------------------------\n\nUse ``cleanlab`` to identify ~100,000 label errors in the 2012 ImageNet training dataset. \n\n.. figure:: https://raw.githubusercontent.com/cgnorthcutt/cleanlab/master/img/imagenet_train_label_errors_32.jpg\n :align: center\n :alt: Image depicting label errors in ImageNet train set \n\nTop label issues in the 2012 ILSVRC ImageNet train set identified using ``cleanlab``. Label Errors are boxed in red. Ontological issues in green. Multi-label images in blue.\n\n``cleanlab``: Find Label Errors in MNIST\n----------------------------------------\n\nUse ``cleanlab`` to identify ~50 label errors in the MNIST dataset. \n\n.. figure:: https://raw.githubusercontent.com/cgnorthcutt/cleanlab/master/img/mnist_training_label_errors24_prune_by_noise_rate.png\n :align: center\n :alt: Image depicting label errors in MNIST train set \n\nLabel errors of the original MNIST **train** dataset identified algorithmically using cleanlab. Depicts the 24 least confident labels, ordered left-right, top-down by increasing self-confidence (probability of belonging to the given label), denoted conf in teal. The label with the largest predicted probability is in green. Overt errors are in red.\n\n\n``cleanlab`` Generality: View performance across 4 distributions and 9 classifiers.\n-----------------------------------------------------------------------------------\n\nUse ``cleanlab`` to learn with noisy labels regardless of dataset distribution or classifier. \n\n.. figure:: https://raw.githubusercontent.com/cgnorthcutt/cleanlab/master/img/demo_cleanlab_across_datasets_and_classifiers.png\n :align: center\n :alt: Image depicting generality of cleanlab across datasets and classifiers \n\nEach sub-figure in the figure above depicts the decision boundary learned using ``cleanlab.classification.LearningWithNoisyLabels`` in the presence of extreme (\\~35%) label errors. Label errors are circled in green. Label noise is class-conditional (not simply uniformly random). Columns are organized by the classifier used, except the left-most column which depicts the ground-truth dataset distribution. Rows are organized by dataset used.\n\nThe code to reproduce this figure is available `here <https://github.com/cgnorthcutt/cleanlab/blob/master/examples/classifier_comparison.ipynb>`__.\n\nEach figure depicts accuracy scores on a test set as decimal values: \n\n1. LEFT (in black): The classifier test accuracy trained with perfect labels (no label errors). \n2. MIDDLE (in blue): The classifier test accuracy trained with noisy labels using ``cleanlab``. \n3. RIGHT (in white): The baseline classifier test accuracy trained with noisy labels.\n\nAs an example, this is the noise matrix (noisy channel) *P(s \\| y)* characterizing the label noise for the first dataset row in the figure. *s* represents the observed noisy labels and *y* represents the latent, true labels. The trace of this matrix is 2.6. A trace of 4 implies no label noise. A cell in this matrix is read like, \"A random 38% of '3' labels were flipped to '2' labels.\"\n\n====== ==== ==== ==== ==== \np(s|y) y=0 y=1 y=2 y=3\n====== ==== ==== ==== ==== \ns=0 0.55 0.01 0.07 0.06\ns=1 0.22 0.87 0.24 0.02\ns=2 0.12 0.04 0.64 0.38\ns=3 0.11 0.08 0.05 0.54\n====== ==== ==== ==== ====\n\n\nGet started with easy, quick examples.\n======================================\n\nNew to **cleanlab**? Start with:\n\n1. `Visualizing confident\n learning <https://github.com/cgnorthcutt/cleanlab/blob/master/examples/visualizing_confident_learning.ipynb>`__\n2. `A simple example of learning with noisy labels on the multiclass\n Iris dataset <https://github.com/cgnorthcutt/cleanlab/blob/master/examples/iris_simple_example.ipynb>`__.\n\nThese examples show how easy it is to characterize label noise in\ndatasets, learn with noisy labels, identify label errors, estimate\nlatent priors and noisy channels, and more.\n\n.. ..\n\n <!---\n\n\n\n ![Image depicting label errors in MNIST test set.](https://raw.githubusercontent.com/cgnorthcutt/cleanlab/master/img/mnist_test_label_errors8.png)\n Selected label errors in the MNIST **test** dataset ordered by increasing self-confidence (in teal).\n\n ## Automatically identify ~5k (of 50k) validation set label errors in ImageNet. [[link]](examples/finding_ImageNet_label_errors).\n ![Image depicting label errors in ImageNet validation set.](https://raw.githubusercontent.com/cgnorthcutt/cleanlab/master/img/imagenet_validation_label_errors_96_prune_by_noise_rate.jpg)\n Label errors in the 2012 ImageNet validation dataset identified automatically with cleanlab using a pre-trained resnet18. Displayed are the 96 least confident labels. We see that ImageNet contains numerous multi-label images, although it is used widely by the machine learning and vision communities as a single-label benchmark dataset.\n\n --->\n\nUse ``cleanlab`` with any model (Tensorflow, caffe2, PyTorch, etc.)\n-------------------------------------------------------------------\n\nAll of the features of the ``cleanlab`` package work with **any model**.\nYes, any model. Feel free to use PyTorch, Tensorflow, caffe2,\nscikit-learn, mxnet, etc. If you use a scikit-learn classifier, all\n``cleanlab`` methods will work out-of-the-box. It\u2019s also easy to use\nyour favorite model from a non-scikit-learn package, just wrap your\nmodel into a Python class that inherits the\n``sklearn.base.BaseEstimator``:\n\n.. code:: python\n\n from sklearn.base import BaseEstimator\n class YourFavoriteModel(BaseEstimator): # Inherits sklearn base classifier\n def __init__(self, ):\n pass\n def fit(self, X, y, sample_weight=None):\n pass\n def predict(self, X):\n pass\n def predict_proba(self, X):\n pass\n def score(self, X, y, sample_weight=None):\n pass\n\n # Now you can use your model with `cleanlab`. Here's one example:\n from cleanlab.classification import LearningWithNoisyLabels\n lnl = LearningWithNoisyLabels(clf=YourFavoriteModel())\n lnl.fit(train_data, train_labels_with_errors)\n\nWant to see a working example? `Here\u2019s a compliant PyTorch MNIST CNN class <https://github.com/cgnorthcutt/cleanlab/blob/master/cleanlab/models/mnist_pytorch.py#L28>`__\n^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n\nAs you can see\n`here <https://github.com/cgnorthcutt/cleanlab/blob/master/cleanlab/models/mnist_pytorch.py#L28>`__,\ntechnically you don\u2019t actually need to inherit from\n``sklearn.base.BaseEstimator``, as you can just create a class that\ndefines .fit(), .predict(), and .predict_proba(), but inheriting makes\ndownstream scikit-learn applications like hyper-parameter optimization\nwork seamlessly. For example, the `LearningWithNoisyLabels()\nmodel <https://github.com/cgnorthcutt/cleanlab/blob/master/cleanlab/classification.py#L48>`__\nis fully compliant.\n\nNote, some libraries exists to do this for you. For pyTorch, check out\nthe ``skorch`` Python library which will wrap your ``pytorch`` model\ninto a ``scikit-learn`` compliant model.\n\n\nDocumentation by Example\n========================\n\n``cleanlab`` Core Package Components\n------------------------------------\n\n1. **cleanlab/classification.py** - The LearningWithNoisyLabels() class for learning with noisy labels.\n2. **cleanlab/latent_algebra.py** -\tEqualities when noise information is known.\n3. **cleanlab/latent_estimation.py** -\tEstimates and fully characterizes all variants of label noise.\n4. **cleanlab/noise_generation.py** - Generate mathematically valid synthetic noise matrices.\n5. **cleanlab/polyplex.py** -\tCharacterizes joint distribution of label noise EXACTLY from noisy channel.\n6. **cleanlab/pruning.py** - Finds the indices of the examples with label errors in a dataset.\n\nMany of these methods have default parameters that won\u2019t be covered\nhere. Check out the method docstrings for full documentation.\n\n\nMethods to Standardize Research with Noisy Labels\n-------------------------------------------------\n\n``cleanlab`` supports a number of functions to generate noise for benchmarking and standardization in research. This next example shows how to generate valid, class-conditional, unformly random noisy channel matrices:\n\n.. code:: python\n\n # Generate a valid (necessary conditions for learnability are met) noise matrix for any trace > 1\n from cleanlab.noise_generation import generate_noise_matrix_from_trace\n noise_matrix=generate_noise_matrix_from_trace(\n K=number_of_classes, \n trace=float_value_greater_than_1_and_leq_K,\n py=prior_of_y_actual_labels_which_is_just_an_array_of_length_K,\n frac_zero_noise_rates=float_from_0_to_1_controlling_sparsity,\n )\n\n # Check if a noise matrix is valid (necessary conditions for learnability are met)\n from cleanlab.noise_generation import noise_matrix_is_valid\n is_valid=noise_matrix_is_valid(noise_matrix, prior_of_y_which_is_just_an_array_of_length_K)\n\nFor a given noise matrix, this example shows how to generate noisy labels. Methods can be seeded for reproducibility.\n\n.. code:: python\n\n # Generate noisy labels using the noise_marix. Guarantees exact amount of noise in labels.\n from cleanlab.noise_generation import generate_noisy_labels\n s_noisy_labels = generate_noisy_labels(y_hidden_actual_labels, noise_matrix)\n\n # This package is a full of other useful methods for learning with noisy labels.\n # The tutorial stops here, but you don't have to. Inspect method docstrings for full docs.\n\n\nEstimate the confident joint, the latent noisy channel matrix, *P(s \\| y)* and inverse, *P(y \\| s)*, the latent prior of the unobserved, actual true labels, *p(y)*, and the predicted probabilities.\n------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------\n\n*s* denotes a random variable that represents the observed, noisy\nlabel and *y* denotes a random variable representing the hidden, actual\nlabels. Both *s* and *y* take any of the m classes as values. The\n``cleanlab`` package supports different levels of granularity for\ncomputation depending on the needs of the user. Because of this, we\nsupport multiple alternatives, all no more than a few lines, to estimate\nthese latent distribution arrays, enabling the user to reduce\ncomputation time by only computing what they need to compute, as seen in\nthe examples below.\n\nThroughout these examples, you\u2019ll see a variable called\n*confident_joint*. The confident joint is an m x m matrix (m is the\nnumber of classes) that counts, for every observed, noisy class, the\nnumber of examples that confidently belong to every latent, hidden\nclass. It counts the number of examples that we are confident are\nlabeled correctly or incorrectly for every pair of obseved and\nunobserved classes. The confident joint is an unnormalized estimate of\nthe complete-information latent joint distribution, *Ps,y*. Most of the\nmethods in the **cleanlab** package start by first estimating the\n*confident_joint*. You can learn more about this in the `confident learning paper <https://arxiv.org/abs/1911.00068>`__.\n\nOption 1: Compute the confident joint and predicted probs first. Stop if that\u2019s all you need.\n^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n\n.. code:: python\n\n from cleanlab.latent_estimation import estimate_latent\n from cleanlab.latent_estimation import estimate_confident_joint_and_cv_pred_proba\n\n # Compute the confident joint and the n x m predicted probabilities matrix (psx),\n # for n examples, m classes. Stop here if all you need is the confident joint.\n confident_joint, psx = estimate_confident_joint_and_cv_pred_proba(\n X=X_train, \n s=train_labels_with_errors,\n clf=logreg(), # default, you can use any classifier\n )\n\n # Estimate latent distributions: p(y) as est_py, P(s|y) as est_nm, and P(y|s) as est_inv\n est_py, est_nm, est_inv = estimate_latent(confident_joint, s=train_labels_with_errors)\n\nOption 2: Estimate the latent distribution matrices in a single line of code.\n^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n\n.. code:: python\n\n from cleanlab.latent_estimation import estimate_py_noise_matrices_and_cv_pred_proba\n est_py, est_nm, est_inv, confident_joint, psx = estimate_py_noise_matrices_and_cv_pred_proba(\n X=X_train,\n s=train_labels_with_errors,\n )\n\nOption 3: Skip computing the predicted probabilities if you already have them.\n^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n\n.. code:: python\n\n # Already have psx? (n x m matrix of predicted probabilities)\n # For example, you might get them from a pre-trained model (like resnet on ImageNet)\n # With the cleanlab package, you estimate directly with psx.\n from cleanlab.latent_estimation import estimate_py_and_noise_matrices_from_probabilities\n est_py, est_nm, est_inv, confident_joint = estimate_py_and_noise_matrices_from_probabilities(\n s=train_labels_with_errors, \n psx=psx,\n )\n\n\nCompletely characterize label noise in a dataset:\n-------------------------------------------------\n\nThe joint probability distribution of noisy and true labels, *P(s,y)*, completely characterizes label noise with a class-conditional *m x m* matrix. \n\n.. code:: python\n\n from cleanlab.latent_estimation import estimate_joint\n joint = estimate_joint(\n s=noisy_labels,\n psx=probabilities,\n confident_joint=None, # Provide if you have it already\n )\n\n\nPU learning with cleanlab:\n--------------------------\n\nPU learning is a special case when one of your classes has no error. P stands for the positive class and **is assumed to have zero label errors** and U stands for unlabeled data, but in practice, we just assume the U class is a noisy negative class that contains some positive examples. Thus, the goal of PU learning is to (1) estimate the proportion of positives in the negative class (see `fraction_noise_in_unlabeled_class` in the last example), (2) find the errors (see last example), and (3) train on clean data (see first example below). `cleanlab` does all three, taking into account that there is no label errors in whichever class you specify.\n\nThere are two ways to use `cleanlab` for PU learning. We'll look at each here.\n\nMethod 1. If you are using the cleanlab classifier `LearningWithNoisyLabels()`, and your dataset has exactly two classes (positive = 1, and negative = 0), PU learning is supported directly in `cleanlab`. You can perform PU learning like this:\n\n.. code:: python\n\n from cleanlab.classification import LearningWithNoisyLabels\n from sklearn.linear_model import LogisticRegression\n # Wrap around any classifier. Yup, you can use sklearn/pyTorch/Tensorflow/FastText/etc.\n pu_class = 0 # Should be 0 or 1. Label of class with NO ERRORS. (e.g., P class in PU)\n lnl = LearningWithNoisyLabels(clf=LogisticRegression(), pulearning=pu_class)\n lnl.fit(X=X_train_data, s=train_noisy_labels)\n # Estimate the predictions you would have gotten by training with *no* label errors.\n predicted_test_labels = lnl.predict(X_test)\n\n\nMethod 2. However, you might be using a more complicated classifier that doesn't work well with LearningWithNoisyLabels (see this example for CIFAR-10). Or you might have 3 or more classes. Here's how to use cleanlab for PU learning in this situation.\nTo let cleanlab know which class has no error (in standard PU learning, this is the P class), you need to set the threshold for that class to 1 (1 means the probability that the labels of that class are correct is 1, i.e. that class has no error). Here's the code:\n\n.. code:: python\n\n import numpy as np\n # K is the number of classes in your dataset\n # psx are the cross-validated predicted probabilities.\n # s is the array/list/iterable of noisy labels\n # pu_class is a 0-based integer for the class that has no label errors.\n thresholds = np.asarray([np.mean(psx[:, k][s == k]) for k in range(K)])\n thresholds[pu_class] = 1.0\n\n\nNow you can use cleanlab however you were before.\nJust be sure to pass in this thresholds parameter wherever it applies. For example:\n\n.. code:: python\n\n # Uncertainty quantification (characterize the label noise\n # by estimating the joint distribution of noisy and true labels)\n cj = compute_confident_joint(s, psx, thresholds=thresholds, )\n # Now the noise (cj) has been estimated taking into account that some class(es) have no error.\n # We can use cj to find label errors like this:\n indices_of_label_errors = get_noise_indices(s, psx, confident_joint=cj, )\n\n # In addition to label errors, we can find the fraction of noise in the unlabeled class.\n # First we need the inv_noise_matrix which contains P(y|s) (proportion of mislabeling).\n _, _, inv_noise_matrix = estimate_latent(confident_joint=cj, s=s, )\n # Because inv_noise_matrix contains P(y|s), p (y = anything | s = pu_class) should be 0\n # because the prob(true label is something else | example is in pu_class) is 0.\n # What's more interesting is p(y = anything | s is not put_class), or in the binary case\n # this translates to p(y = pu_class | s = 1 - pu_class) because pu_class is 0 or 1.\n # So, to find the fraction_noise_in_unlabeled_class, for binary, you just compute:\n fraction_noise_in_unlabeled_class = inv_noise_matrix[pu_class][1 - pu_class] \n\n\nNow that you have `indices_of_label_errors`, you can remove those label errors and train on clean data (or only remove some of the label errors and iteratively use confident learning / cleanlab to improve results)\n\n\nThe Polyplex\n------------\n\nThe key to learning in the presence of label errors is estimating the joint distribution between the actual, hidden labels \u2018*y*\u2019 and the observed, noisy labels \u2018*s*\u2019. Using ``cleanlab`` and the theory of confident learning, we can completely characterize the trace of the latent joint distribution, *trace(P(s,y))*, given *p(y)*, for any fraction of label errors, i.e.\u00a0for any trace of the noisy channel, *trace(P(s|y))*.\n\nYou can check out how to do this yourself here: 1. `Drawing\nPolyplices <https://github.com/cgnorthcutt/cleanlab/blob/master/examples/drawing_polyplices.ipynb>`__ 2. `Computing\nPolyplices <https://github.com/cgnorthcutt/cleanlab/blob/master/cleanlab/polyplex.py>`__\n\n\nJoin our community\n------------------\n\nHave ideas for the future of cleanlab? How are you using cleanlab? `Join the discussion <https://github.com/cgnorthcutt/cleanlab/discussions>`__.\n\nHave code improvements for cleanlab? `Submit a code pull request <https://github.com/cgnorthcutt/cleanlab/issues/new>`__.\n\nDo you have an issue with cleanlab? `Submit an issue <https://github.com/cgnorthcutt/cleanlab/issues/new>`__.\n\n\nLicense\n-------\n\nCopyright (c) 2017-2050 Curtis G. Northcutt\n\ncleanlab is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.\n\ncleanlab is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. \n\nSee `GNU General Public LICENSE <https://github.com/cgnorthcutt/cleanlab/blob/master/LICENSE>`__ for details.\n\nTHIS LICENSE APPLIES TO THIS VERSION AND ALL PREVIOUS VERSIONS OF cleanlab.\n\n\n", "bugtrack_url": null, "license": "GPLv3+", "summary": "The standard package for machine learning with noisy labels and finding mislabeled data in Python.", "version": "1.0", "split_keywords": [ "machine_learning", "denoising", "classification", "weak_supervision", "learning_with_noisy_labels", "unsupervised_learning" ], "urls": [ { "comment_text": "", "digests": { "md5": "7cb6a162caa11f2a11e4654ee39040a4", "sha256": "db9119f5ebe536b00748aadf257ccffde26d181ae3e3b731a852b33574ec2f9e" }, "downloads": -1, "filename": "cleanlab-1.0-py2.py3-none-any.whl", "has_sig": false, "md5_digest": "7cb6a162caa11f2a11e4654ee39040a4", "packagetype": "bdist_wheel", "python_version": "py2.py3", "requires_python": ">=2.7,!=3.0.*,!=3.1.*,!=3.2.*,!=3.3.*", "size": 77551, "upload_time": "2021-04-18T19:42:42", "upload_time_iso_8601": "2021-04-18T19:42:42.481297Z", "url": "https://files.pythonhosted.org/packages/c1/8c/4648c2e0337d6634c07d587ed7f954e36c340c581e48c9ad1414c517014e/cleanlab-1.0-py2.py3-none-any.whl", "yanked": false, "yanked_reason": null }, { "comment_text": "", "digests": { "md5": "22a19af3a7a7968dd97075c06f9138f5", "sha256": "0c9ec2e2dc2f46b9aaaa46d8b4f0f2c1c080f15997dc1a6cf517fedbe19773d9" }, "downloads": -1, "filename": "cleanlab-1.0.tar.gz", "has_sig": false, "md5_digest": "22a19af3a7a7968dd97075c06f9138f5", "packagetype": "sdist", "python_version": "source", "requires_python": ">=2.7,!=3.0.*,!=3.1.*,!=3.2.*,!=3.3.*", "size": 125327, "upload_time": "2021-04-18T19:42:44", "upload_time_iso_8601": "2021-04-18T19:42:44.339499Z", "url": "https://files.pythonhosted.org/packages/82/56/9f4c5bc333bff69b56a1bcbbee73e751b0fd997b7960e089fecbdaa79726/cleanlab-1.0.tar.gz", "yanked": false, "yanked_reason": null } ], "upload_time": "2021-04-18 19:42:44", "github": true, "gitlab": false, "bitbucket": false, "github_user": null, "github_project": "cgnorthcutt", "error": "Could not fetch GitHub repository", "lcname": "cleanlab" }

Elapsed time: 0.25700s