[DOC] Update changelog (#311)

Co-authored-by: Felix Dangel <[email protected]>

changelog.md

@@ -9,17 +9,72 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 ## [1.6.0] - 2023-06-26
 
 With this patch, BackPACK supports `torch==2.x` and deprecates `python3.7`
-([PR](https://github.com/f-dangel/backpack/pull/307)).
+([PR](https://github.com/f-dangel/backpack/pull/307)). BackPACK now supports
+`BCEWithLogitsLoss`, and supports KFAC for N-dimensional (transpose)
+convolutions.
 
 ### Added/New
 - Example use case for `retain_graph`
   ([PR](https://github.com/f-dangel/backpack/pull/302))
+- Example for computing batched Jacobians
+  ([PR](https://github.com/f-dangel/backpack/pull/267))
+- Support `BCEWithLogitsLoss`
+  - `BatchDiagHessian` and `DiagHessian` extensions
+    ([PR](https://github.com/f-dangel/backpack/pull/279))
+  - `BatchDiagGGN` (exact, MC) and `DiagGGN` (exact, MC) extensions
+    ([PR](https://github.com/f-dangel/backpack/pull/280))
+  - `KFAC` extension
+    ([PR](https://github.com/f-dangel/backpack/pull/283))
+  - `KFLR` extension
+    ([PR](https://github.com/f-dangel/backpack/pull/284))
+- Extend `KFAC, KFLR, KFRA` to N-dimensional convolutions (and transpose
+  convolutions) ([blog post](https://fdangel.com/posts/kfac_explained.html))
+  - Support for `Conv{1,2,3}d`
+    ([PR](https://github.com/f-dangel/backpack/pull/291))
+  - Support for `ConvTranspose{1,2,3}d`
+    ([PR](https://github.com/f-dangel/backpack/pull/292))
+
+### Fixed/Removed
+- Use correct imports on website code samples
+  ([PR](https://github.com/f-dangel/backpack/pull/262))
 
 ### Internal
 - Update code to latest `black` and `flake8`
   ([PR](https://github.com/f-dangel/backpack/pull/301))
 - Fix examples on splash page
   ([PR](https://github.com/f-dangel/backpack/pull/298))
+- Abstract negative log-likelihood losses in core
+  ([PR](https://github.com/f-dangel/backpack/pull/252))
+  - Apply to `MSELoss` derivatives
+    ([PR](https://github.com/f-dangel/backpack/pull/252))
+  - Apply to `CrossEntropyLoss` derivatives
+    ([PR](https://github.com/f-dangel/backpack/pull/256))
+  - Implement `BCEWithLogitsLoss` derivatives
+    ([PR](https://github.com/f-dangel/backpack/pull/257))
+- Implement second-order derivatives of `BCEWithLogitsLoss`
+  ([PR](https://github.com/f-dangel/backpack/pull/271))
+- Implement optimized sampled gradients for `BCEWithLogitsLoss`
+  ([PR](https://github.com/f-dangel/backpack/pull/278))
+- Use batch size 1 in KFAC tests for ResNets
+  ([PR](https://github.com/f-dangel/backpack/pull/265))
+- Test with PyTorch `1.9.0` and `1.12.0` and make tests compatible
+  ([PR](https://github.com/f-dangel/backpack/pull/277))
+- Use `unfoldNd` package
+  - For input unfolding of convolutions
+    ([PR](https://github.com/f-dangel/backpack/pull/285))
+  - For input unfolding of transpose convolutions
+    ([PR](https://github.com/f-dangel/backpack/pull/287))
+- Fully-document utility functions for convolutions
+  ([PR](https://github.com/f-dangel/backpack/pull/286))
+- Make output shape of unfolded input of transpose convolution consistent with
+  convolution case
+  ([PR](https://github.com/f-dangel/backpack/pull/289))
+- Fully-document `HBPConv2d` class
+  ([PR](https://github.com/f-dangel/backpack/pull/290))
+- Fix support for PyTorch 1.13
+  ([PR](https://github.com/f-dangel/backpack/pull/296))
+- Update linting and formatting to latest `black`, `flake8`
+  ([PR](https://github.com/f-dangel/backpack/pull/301))
 
 ## [1.5.2] - 2022-12-19