Release notes for 1.4.0-preview2 and 0.16.0-preview2. (#4315)

* release notes.

* Release notes for 1.4.0-preview2 and 0.16.0-preview2.

Author: codemzs

docs/release-notes/1.4.0-preview2/release-1.4.0-preview2.md

@@ -0,0 +1,58 @@
+# [ML.NET](http://dot.net/ml) 1.4.0-preview2
+
+## **New Features**
+- **Deep Neural Networks Training (0.16.0-preview2)**
+
+  Improves the in-preview `ImageClassification` API further:
+  - Early stopping feature stops the training when optimal accuracy is reached ([#4237](https://github.com/dotnet/machinelearning/pull/4237))
+  - Enables inferencing on in-memory images ([#4242](https://github.com/dotnet/machinelearning/pull/4242))
+  - `PredictedLabel` output column now contains actual class labels instead of `uint32` class index values ([#4228](https://github.com/dotnet/machinelearning/pull/4228))
+  - GPU support on Windows and Linux ([#4270](https://github.com/dotnet/machinelearning/pull/4270), [#4277](https://github.com/dotnet/machinelearning/pull/4277))
+  - Upgraded [TensorFlow .NET](https://github.com/SciSharp/TensorFlow.NET) version to 0.11.3 ([#4205](https://github.com/dotnet/machinelearning/pull/4205))
+
+  [In-memory image inferencing sample](https://github.com/dotnet/machinelearning/blob/master/docs/samples/Microsoft.ML.Samples/Dynamic/ImageClassification/ResnetV2101TransferLearningTrainTestSplit.cs)  
+  [Early stopping sample](https://github.com/dotnet/machinelearning/blob/master/docs/samples/Microsoft.ML.Samples/Dynamic/ImageClassification/ResnetV2101TransferLearningEarlyStopping.cs)  
+  [GPU samples](https://github.com/dotnet/machinelearning/tree/master/docs/samples/Microsoft.ML.Samples.GPU)  
+
+- **New ONNX Exporters (1.4.0-preview2)**
+  - LpNormNormalizing transformer ([#4161](https://github.com/dotnet/machinelearning/pull/4161))
+  - PCA transformer ([4188](https://github.com/dotnet/machinelearning/pull/4188))
+  - TypeConverting transformer ([#4155](https://github.com/dotnet/machinelearning/pull/4155))
+  - MissingValueIndicator transformer ([#4194](https://github.com/dotnet/machinelearning/pull/4194))
+
+## **Bug Fixes**
+- OnnxSequenceType and ColumnName attributes together doesn't work ([#4187](https://github.com/dotnet/machinelearning/pull/4187))
+- Fix memory leak in TensorflowTransformer ([#4223](https://github.com/dotnet/machinelearning/pull/4223))
+- Enable permutation feature importance to be used with model loaded from disk ([#4262](https://github.com/dotnet/machinelearning/pull/4262))
+- `IsSavedModel` returns true when loaded TensorFlow model is a frozen model ([#4262](https://github.com/dotnet/machinelearning/pull/4197))
+- Exception when using `OnnxSequenceType` attribute directly without specify sequence type ([#4272](https://github.com/dotnet/machinelearning/pull/4272), [#4297](https://github.com/dotnet/machinelearning/pull/4297))
+
+## **Samples**
+- TensorFlow full model retrain sample ([#4127](https://github.com/dotnet/machinelearning/pull/4127))
+
+## **Breaking Changes**
+None.
+
+## **Obsolete API**
+- `OnnxSequenceType` attribute that doesn't take a type ([#4272](https://github.com/dotnet/machinelearning/pull/4272), [#4297](https://github.com/dotnet/machinelearning/pull/4297))
+
+## **Enhancements**
+- Improve exception message in LightGBM ([#4214](https://github.com/dotnet/machinelearning/pull/4214))
+- FeaturizeText should allow only outputColumnName to be defined ([#4211](https://github.com/dotnet/machinelearning/pull/4211))
+- Fix NgramExtractingTransformer GetSlotNames to not allocate a new delegate on every invoke ([#4247](https://github.com/dotnet/machinelearning/pull/4247)) 
+- Resurrect broken code coverage build and re-enable code coverage for pull request ([#4261](https://github.com/dotnet/machinelearning/pull/4261))
+- NimbusML entrypoint for permutation feature importance ([#4232](https://github.com/dotnet/machinelearning/pull/4232))
+- Reuse memory when copying outputs from TensorFlow graph ([#4260](https://github.com/dotnet/machinelearning/pull/4260))
+- DateTime to DateTime standard conversion ([#4273](https://github.com/dotnet/machinelearning/pull/4273))
+- CodeCov version upgraded to 1.7.2 ([#4291](https://github.com/dotnet/machinelearning/pull/4291))
+ 
+## **CLI and AutoML API**
+None.
+
+## **Remarks**
+None.
+
+
+
+
+