|
| 1 | +# [ML.NET](http://dot.net/ml) 1.5.0 |
| 2 | + |
| 3 | +## **New Features** |
| 4 | +- **New anomaly detection algorithm** ([#5135](https://github.com/dotnet/machinelearning/pull/5135)). ML.NET has previously supported anomaly detection through [DetectAnomalyBySrCnn](https://docs.microsoft.com/en-us/dotnet/api/microsoft.ml.timeseriescatalog.detectanomalybysrcnn?view=ml-dotnet). This function operates in a streaming manner by computing anomalies around each arriving point and examining a window around it. Now we introduce a new function `DetectEntireAnomalyBySrCnn` that computes anomalies by considering the entire dataset and also supports the ability to set sensitivity and output margin. |
| 5 | +- **Root Cause Detection** ([#4925](https://github.com/dotnet/machinelearning/pull/4925)) ML.NET now also supports root cause detection for anomalies detected in time series data. |
| 6 | + |
| 7 | +## **Enhancements** |
| 8 | +- **Updates to TextLoader** |
| 9 | + - Enable TextLoader to accept new lines in quoted fields ([#5125](https://github.com/dotnet/machinelearning/pull/5125)) |
| 10 | + - Add escapeChar support to TextLoader ([#5147](https://github.com/dotnet/machinelearning/pull/5147)) |
| 11 | + - Add public generic methods to TextLoader catalog that accept Options objects ([#5134](https://github.com/dotnet/machinelearning/pull/5134)) |
| 12 | + - Added decimal marker option in TextLoader ([#5145](https://github.com/dotnet/machinelearning/pull/5145), [#5154](https://github.com/dotnet/machinelearning/pull/5154)) |
| 13 | +- Onnxruntime updated to v1.3 ([#5104](https://github.com/dotnet/machinelearning/pull/5104)). This brings support for additional data types for the HashingEstimator. |
| 14 | +- Onnx export for OneHotHashEncodingTransformer and HashingTransormer ([#5013](https://github.com/dotnet/machinelearning/pull/5013), [#5152](https://github.com/dotnet/machinelearning/pull/5152), [#5138](https://github.com/dotnet/machinelearning/pull/5138)) |
| 15 | +- Support for Categorical features in CalculateFeatureContribution of LightGBM ([#5018](https://github.com/dotnet/machinelearning/pull/5018)) |
| 16 | + |
| 17 | + |
| 18 | +## **Bug Fixes** |
| 19 | +In this release we have traced down every bug that would occur randomly and sporadically and fixed many subtle bugs. As a result, we have also re-enabled a lot of tests listed in the **Test Updates** section below. |
| 20 | +- Fixed race condition for test MulticlassTreeFeaturizedLRTest ([#4950](https://github.com/dotnet/machinelearning/pull/4950)) |
| 21 | +- Fix SsaForecast bug ([#5023](https://github.com/dotnet/machinelearning/pull/5023)) |
| 22 | +- Fixed x86 crash ([#5081](https://github.com/dotnet/machinelearning/pull/5081)) |
| 23 | +- Fixed and added unit tests for EnsureResourceAsync hanging issue ([#4943](https://github.com/dotnet/machinelearning/pull/4943)) |
| 24 | +- Added IDisposable support for several classes ([#4939](https://github.com/dotnet/machinelearning/pull/4939)) |
| 25 | +- Updated libmf and corresponding MatrixFactorizationSimpleTrainAndPredict() baselines per build ([#5121](https://github.com/dotnet/machinelearning/pull/5121)) |
| 26 | +- Fix MatrixFactorization trainer's warning ([#5071](https://github.com/dotnet/machinelearning/pull/5071)) |
| 27 | +- Update CodeGenerator's console project to netcoreapp3.1 ([#5066](https://github.com/dotnet/machinelearning/pull/5066)) |
| 28 | +- Let ImageLoadingTransformer dispose the last image it loads ([#5056](https://github.com/dotnet/machinelearning/pull/5056)) |
| 29 | +- [LightGBM] Fixed bug for empty categorical values ([#5048](https://github.com/dotnet/machinelearning/pull/5048)) |
| 30 | +- Converted potentially large variables to type long ([#5041](https://github.com/dotnet/machinelearning/pull/5041)) |
| 31 | +- Made resource downloading more robust ([#4997](https://github.com/dotnet/machinelearning/pull/4997)) |
| 32 | +- Updated MultiFileSource.Load to fix inconsistent behavior with multiple files ([#5003](https://github.com/dotnet/machinelearning/pull/5003)) |
| 33 | +- Removed WeakReference<IHosts> already cleaned up by GC ([#4995](https://github.com/dotnet/machinelearning/pull/4995)) |
| 34 | +- Fixed Bitmap(file) locking the file. ([#4994](https://github.com/dotnet/machinelearning/pull/4994)) |
| 35 | +- Remove WeakReference list in PredictionEnginePoolPolicy. ([#4992](https://github.com/dotnet/machinelearning/pull/4992)) |
| 36 | +- Added the assembly name of the custom transform to the model file ([#4989](https://github.com/dotnet/machinelearning/pull/4989)) |
| 37 | +- Updated constructor of ImageLoadingTransformer to accept empty imageFolder paths ([#4976](https://github.com/dotnet/machinelearning/pull/4976)) |
| 38 | + |
| 39 | +**Onnx bug fixes** |
| 40 | +- ColumnSelectingTransformer now infers ONNX shape ([#5079](https://github.com/dotnet/machinelearning/pull/5079)) |
| 41 | +- Fixed KMeans scoring differences between ORT and OnnxRunner ([#4942](https://github.com/dotnet/machinelearning/pull/4942)) |
| 42 | +- CountFeatureSelectingEstimator no selection support ([#5000](https://github.com/dotnet/machinelearning/pull/5000)) |
| 43 | +- Fixes OneHotEncoding Issue ([#4974](https://github.com/dotnet/machinelearning/pull/4974)) |
| 44 | +- Fixes multiclass logistic regression ([#4963](https://github.com/dotnet/machinelearning/pull/4963)) |
| 45 | +- Adding vector tests for KeyToValue and ValueToKey ([#5090](https://github.com/dotnet/machinelearning/pull/5090)) |
| 46 | + |
| 47 | +**AutoML fixes** |
| 48 | +- Handle NaN optimization metric in AutoML ([#5031](https://github.com/dotnet/machinelearning/pull/5031)) |
| 49 | +- Add projects capability in CodeGenerator ([#5002](https://github.com/dotnet/machinelearning/pull/5002)) |
| 50 | +- Simplify CodeGen - phase 2 ([#4972](https://github.com/dotnet/machinelearning/pull/4972)) |
| 51 | +- Support sweeping multiline option in AutoML ([#5148](https://github.com/dotnet/machinelearning/pull/5148)) |
| 52 | + |
| 53 | + |
| 54 | +## **Test updates** |
| 55 | +- Fix libomp installation for MacOS Builds([#5143](https://github.com/dotnet/machinelearning/pull/5143), [#5141](https://github.com/dotnet/machinelearning/pull/5141)) |
| 56 | +- address TF test download fail, use resource manager with retry download ([#5102](https://github.com/dotnet/machinelearning/pull/5102)) |
| 57 | +- Adding OneHotHashEncoding Test ([#5098](https://github.com/dotnet/machinelearning/pull/5098)) |
| 58 | +- Changed Dictionary to ConcurrentDictionary ([#5097](https://github.com/dotnet/machinelearning/pull/5097)) |
| 59 | +- Added SQLite database to test loading of datasets in non-Windows builds ([#5080](https://github.com/dotnet/machinelearning/pull/5080)) |
| 60 | +- Added ability to compare configuration specific baselines, updated baslines for many tests and re-enabled disabled tests ([#5045](https://github.com/dotnet/machinelearning/pull/5045), [#5059](https://github.com/dotnet/machinelearning/pull/5059), [#5068](https://github.com/dotnet/machinelearning/pull/5068), [#5057](https://github.com/dotnet/machinelearning/pull/5057), [#5047](https://github.com/dotnet/machinelearning/pull/5047), [#5029](https://github.com/dotnet/machinelearning/pull/5029), [#5094](https://github.com/dotnet/machinelearning/pull/5094), [#5060](https://github.com/dotnet/machinelearning/pull/5060)) |
| 61 | +- Fixed TestCancellation hanging ([#4999](https://github.com/dotnet/machinelearning/pull/4999)) |
| 62 | +- fix benchmark test hanging issue ([#4985](https://github.com/dotnet/machinelearning/pull/4985)) |
| 63 | +- Added working version of checking whether file is available for access ([#4938](https://github.com/dotnet/machinelearning/pull/4938)) |
| 64 | + |
| 65 | +## **Documentation Updates** |
| 66 | +- Update OnnxTransformer Doc XML ([#5085](https://github.com/dotnet/machinelearning/pull/5085)) |
| 67 | +- Updated build docs for .NET Core 3.1 ([#4967](https://github.com/dotnet/machinelearning/pull/4967)) |
| 68 | +- Updated OnnxScoringEstimator's documentation ([#4966](https://github.com/dotnet/machinelearning/pull/4966)) |
| 69 | +- Fix xrefs in the LDSVM trainer docs ([#4940](https://github.com/dotnet/machinelearning/pull/4940)) |
| 70 | +- Clarified parameters on time series ([#5038](https://github.com/dotnet/machinelearning/pull/5038)) |
| 71 | +- Update ForecastBySsa function specifications and add seealso ([#5027](https://github.com/dotnet/machinelearning/pull/5027)) |
| 72 | +- Add see also section to TensorFlowEstimator docs ([#4941](https://github.com/dotnet/machinelearning/pull/4941)) |
| 73 | + |
| 74 | + |
| 75 | +## **Breaking Changes** |
| 76 | +- None |
| 77 | + |
| 78 | + |
| 79 | + |
0 commit comments