- [Feature] Removes telemetry
- [Feature] Adds
ploomber-extension
as dependency - [Feature] Support for mistune v3, altair v5
- [Doc] Added example for
.upsert_append()
- [Feature] Adds coloring scheme and style to plots (#243)
- [Feature] Adds new report API for evaluating and comparing models
- [Doc] Tutorial covering underfitting/overfitting
- [Doc] User guide on Feature Selection
- [Feature] Adds
plot.InteractiveConfusionMatrix
object-oriented API (#275) - [Doc] User guide on interactive confusion matrix plot (#275)
- [Doc] Reorganisation of the classification module (#297)
- [Doc] Added API for RandomForestClassifierGrid (#297)
- [Fix] Adds input validation with suggestions to
plot.grid_search
(#53) - [Fix] Fixes an error that prevented
ClassificationReport.from_raw_data
from displaying the plot - [Doc] Object oriented API guide includes appropriate methods under each model (#271)
- [Doc] Quickstart clustering guide now uses
plot.pca
(#276) - [Doc] User Guide on micro and macro averaging
- [Feature] Improves composed ROC plots so that different curves are more easily identifiable
- [Feature] Adds
plot.CalibrationCurve
object-oriented API - [Doc] User Guide on Probability Calibration
- [Doc] User Guide on real-time tracking
- [API Change]
ax
parameter removed from ROC constructor (#255) - [API Change] Applies deprecation to
plot.ClassificationReport
- [Feature] Adds
plot.PrecisionRecall
object-oriented API [#189] - [Feature] Adds
plot.pca
- [Doc] User guide on PCA
- [API Change] A new figure and axes is created (via
plt.subplots()
) when calling a plotting method withax=None
. Previously, the current axes was used (viaplt.gca()
) (#211) - [API Change]
silhouette_analysis_from_results
function now accepts a list of cluster labels #213 - [API Change] Removed
n_clusters
fromelbow_curve
(#247) - [Feature] Adds
RandomForestClassifierGrid
- [Fix] Validates
plot.ROC
inputs (#98) - [Fix] Uses existing labels in multi ROC curve (#225)
- [Fix] Validating input elbow curve model has "score" method [#146]
- [Fix] Adds class labels for multi class roc plot (#209)
- [Fix] Fixed ROC curves plots to show properly.
- [Fix]
setup.py
fix due to change in setuptools 67.0.0
- [API Change] Applies deprecation to ROC constructor (#135)
- [API Change] Applies deprecation to ConfusionMatrix constructor
- [API Change] Deprecates
n_clusters
inplot.elbow_curve
in favor ofrange_n_clusters
(#148) - [Fix] Adds labels to
plot.elbow_curve
(#152) - [Fix] Showing community link when throwing exceptions
- Adds support for
pandas.Series
to prediction error plot (#154) - Fixes error to prediction plot due to hardcoded axis (#155)
- Deprecates
model
argument in prediction plot
- Fixes to
plot.elbow_curve
: no longer hardcoding figure size, random seed for model, also showingn_clusters
in title - Fixes error when passing array to
table.feature_importances
(#144) - Adds clustering quick start
- New
plot.ROC
api added - Adds
plot.ROC.__add__
for generating overlapping curves plot.ROC
can be serialized/unserialized with.dump
/.from_dump
- Adds feature ranking plots
- Adds
plot.ks_statistic
- Adds
plot.cumulative_gain
- Adds
plot.lift_curve
- Adds
plot.cooks_distance
- Adds
plot.silhoutte_analysis
- Adds clustering user guide
- Adds
plot.target_analysis
- Adds bulk insert to
SQliteTracker
via.insert_many
SQliteTracker.{get_parameters_keys, get_sample_query}
support extracting keys from nested JSON objects
plot.ConfusionMatrix
andplot.ClassifictionReport
can be serialized/unserialized with.dump
/.from_dump
- Adds
Experiment
class to easily create new experiments fromSQLiteTracker
using.new_experiment()
- Adds
Experiment.log_classification_report
- Adds
Experiment.log_confusion_matrix
- Adds
Experiment.log_figure
- Adds
Experiment.log
- Adds
Experiment.log_dict
- Adds
SQLiteTracker.get
- Adds docstring examples to
SQLiteTracker
andExperiment
- Updates
SQliteTracker
tutorial - Fixes error when querying experiments using
SQliteTracker
whose UUID was all numbers (e.g.,"1234"
)
- Adds
plot.residuals
andplot.prediction_error
for evaluating regression models
SQliteTracker.get_sample_query
generates a query compatible with older SQLite versions that do not support the->
operatorSQliteTracker
creates shorter experiment IDs- Fixes whitespace in generated HTML when generating tab views and in
SQLTracker
results whenas_frame=False
- Adds
as_frame
argument toSQLiteTracker
- Adds
SQLiteTracker.upsert
- Allow overwriting records in
SQLiteTracker.update
withallow_overwrite=True
- Adds
SQliteTracker.get_sample_query
- Adds
SQliteTracker.get_parameters_keys
- Adds
plot.scores_distribution
- Adds
plot.classification_report
- Fixes
plot.calibration_curve
error that scaled probabilities before plotting
plot.calibration_curve
allows plotting curves with different sample sizes
- Adds
plot.calibration_curve
- Renames
cluster_ranges
ton_clusters
inplot.elbow_curve
- Adds
plot.elbow_curve_from_results
- Adds
plot.elbow_curve
- Updates telemetry
- Adds [plot.ConfusionMatrix]{.title-ref}
- Updates telemetry key
NotebookDatabase
: makespath
the primary keyNotebookDatabase
:.index()
uses path to see if the notebook has been indexedNotebookDatabase
: addsupdate
to.index()
- Adds anonymous telemetry
- Query notebooks with SQL using
NotebookDatabase
- Stripping output string in
NotebookIntrospector
- Ignoring standard error output in
NotebookIntrospector
- Adds
sort
argument toplot.grid_search
(#45)
- Fixes an error in
plot.grid_search
when parameters grid has a single parameter
- Adds
NotebookIntrospector.get_injected_parameters
- Fixes error that caused grid search plot to fail when a parameter
had a
None
value (#40)
- Adds missing dependency (
IPython
), required byNotebookIntrospector
NotebookCollection
displays output using HTML and tabs- Adds links to try out the examples in binder
DataSelector
copies input steps to prevent mutating input params- Simplifies
NotebookInstrospector
API and adds first implementation ofNotebookCollection
- Adds SQLiteTracker for tracking ML experiments using a SQlite backend
- Adds NotebookIntrospector [Experimental]
- Migrates tests to
nox
- Adds DataSelector
- Enables testing with Python 3.8
- Drops support for Python 3.5
- Documentation migrated to Read the Docs