oneDAL is happy to introduce 2025.10.1 release!

What's Changed

• The release serves as a support release for Extension for Scikit-learn* 2025.10.1

Thanks to everyone who helped us make 2025.10.1 release possible!

@Alexandr-Solovev, @Vika-F, @david-cortes-intel, @icfaust, @napetrov, @maria-Petrova, @homksei, @ahuber21, @ethanglaser, @razdoburdin, @avolkov-intel, @yuejiaointel, @DDJHB, @kjackiew, @richardnorth3

oneDAL is happy to introduce 2025.10.0 release!

🚩 Removals and ABI Compatibility

• The following functionality is deprecated and is to be removed in the oneDAL 2026.0.0 release:
  • Matrix, PackedSymmetricMatrix and PackedTriangularMatrix data types
  • ODBC, KDB and MySQL data sources
  • The algorithms listed in the 20251015-DAAL-algorithms-deprecation RFC will be fully deprecated
  • The algorithms listed in the 20251022-DAAL-interfaces-deprecation RFC will be removed from the DAAL API, but their functionality will be left available through other APIs

🚨 What's New

• Introduced new oneDAL functionality:
  • Vectorized forest mean/variance calculations

🪲 Bug Fixes

• Fixed makefile incorrect flags for VC compiler
• Fixed divisions by zero when there are weights of zero in forests
• Fixed memory leaks in DAAL
• Fixed vectorization pragmas for icx, GNU, clang and MS VC compilers
• Fix undefined behavior warning with GCC

Acknowledgements

Thanks to everyone who helped us make 2025.10.0 release possible!

@Alexsandruss, @Alexandr-Solovev, @Vika-F, @david-cortes-intel, @icfaust, @napetrov, @maria-Petrova, @homksei, @ahuber21, @ethanglaser, @razdoburdin, @avolkov-intel, @yuejiaointel, @DDJHB, @kjackiew, @richardnorth3

New Contributors

• @avineet4 made their first contribution in #3410

Full Changelog: 2025.9.0...2025.10.0

oneDAL is happy to introduce 2025.9.0 release!

🚨 What's New

• Introduced new oneDAL functionality:
  • Enabled SPMD API support in kNN search

🪲 Bug Fixes

• Fixed insufficient memory allocation in csr_table conversion
• Prevented Logistic Regression from turning to inf in predictions

Acknowledgements

Thanks to everyone who helped us make 2025.9.0 release possible!

@Alexsandruss, @Alexandr-Solovev, @Vika-F, @david-cortes-intel, @icfaust, @napetrov, @maria-Petrova, @homksei, @ahuber21, @ethanglaser, @razdoburdin, @avolkov-intel, @KateBlueSky, @yuejiaointel, @DDJHB, @kjackiew, @richardnorth3

Full Changelog: 2025.8.0...2025.9.0

oneDAL is happy to introduce 2025.8.0 release!

🚨 What's New

• Introduced new oneDAL functionality:
  • Improved numerical precision of logistic function
  • Added cosine distance algorithm
  • Added SLEEF-based SVE-optimized float exponential implementation for postGemmPart function
  • Improved performance of kNN prediction

🪲 Bug Fixes

• Fixed illegal instruction on SSE4.2
• Fixed error in modelbuilders examples
• Logistic loss now avoids turning to infinite

Acknowledgements

Thanks to everyone who helped us make 2025.8.0 release possible!

@Alexsandruss, @Alexandr-Solovev, @Vika-F, @david-cortes-intel, @icfaust, @napetrov, @maria-Petrova, @homksei, @ahuber21, @ethanglaser, @razdoburdin, @avolkov-intel, @KateBlueSky, @yuejiaointel, @DDJHB, @kjackiew, @richardnorth3

New Contributors

• @shubhamsvc made their first contribution in #3271

Full Changelog: 2025.7.0...2025.8.0

oneDAL is happy to introduce 2025.7.0 release!

🚨 What's New

• Introduced new oneDAL functionality:
  • Improved thread safety in dynamic symbols loading on Windows
  • Added bazel support for daal examples
  • Added correlation distance algorithm
  • Added grain_size hyperparameter into Covariance and PCA algorithms

🪲 Bug Fixes

• Fixed shared library AddressSanitizer integration for the gnu compiler
• Fixed the xcsrmv function for correct covariance algorithm computation
• Fixed hyperparameterIdCount in Covariance

Acknowledgements

Thanks to everyone who helped us make 2025.7.0 release possible!

@Alexsandruss, @Alexandr-Solovev, @Vika-F, @david-cortes-intel, @icfaust, @napetrov, @maria-Petrova, @homksei, @ahuber21, @ethanglaser, @razdoburdin, @avolkov-intel

Full Changelog: 2025.6.0...2025.7.0

oneDAL is happy to introduce 2025.6.1 release!

🚩 Removals and ABI Compatibility

• The oneDAL static SYCL library is deprecated and removed in the oneDAL 2025.6.1 release

🔨 Library Engineering

• Replaced CPATH variable with CPLUS_INCLUDE_PATH

🚨 What's New

• Introduced new oneDAL functionality:
  • Added internal support for logging, tracing, and analysis
  • Enabled the use of reduction primitives based on tbb::parallel_reduce and tbb::parallel_deterministic_reduce
  • Improved prediction and modeling performance through optimized linear regression
  • Higher throughput for vector processing onall x86 CPUs by enabling higher bit SIMD instructions
  • Improved precision for distance measurement thru distance correlation
  • Improved stability for large row datasets on GPU

🪲 Bug Fixes

• Fixed issues with Logistic Regression and Newton-CG

Acknowledgements

Thanks to everyone who helped us make 2025.6.1 release possible!

@Alexsandruss, @Alexandr-Solovev, @Vika-F, @david-cortes-intel, @icfaust, @napetrov, @maria-Petrova, @homksei, @ahuber21, @ethanglaser, @razdoburdin, @avolkov-intel

Full Changelog: 2025.5.0...2025.6.0

oneDAL is happy to introduce 2025.5.0 release!

🚩 Removals and ABI Compatibility

• Removed dispatching to SSE2 code path on non-intel x86 CPUs

🚨 What's New

• Introduced new oneDAL functionality:
  • Added new parameters for linear regression
  • Introduced non-batched linear regression for high-dimensional problems
  • Added hyperparameters support into Decision Forest training algorithm on CPU
  • Added SVE implementation of sumWithSIMD
  • Added float datatype support to rocAucScore
  • Optimized Random Forest algorithm

🪲 Bug Fixes

• Fixed sparse K-means failure on PVC when the number of rows in large
• Fixed potential non-definitiness of BFGS approximation
• Added isValid() method for dtrees ModelImpl

Acknowledgements

Thanks to everyone who helped us make 2025.5.0 release possible!

@Alexsandruss, @Alexandr-Solovev, @Vika-F, @david-cortes-intel, @icfaust, @napetrov, @maria-Petrova, @homksei, @ahuber21, @ethanglaser, @razdoburdin, @avolkov-intel

New Contributors

• @isuruf made their first contribution in #3026
• @richardnorth3 made their first contribution in #3059
• @yuejiaointel made their first contribution in #3108
• @theComputeKid made their first contribution in #3100

Full Changelog: 2025.4.0...2025.5.0

oneDAL is happy to introduce 2025.4.0 release!

🚨 What's New

• Introduced new oneDAL functionality:
  • Updates to the unified API for oneDAL RNG primitive functions
  • Changed stream handling in the RNG clone

🪲 Bug Fixes

• Fixed issues with the DAAL code samples (building with C++20)

Acknowledgements

Thanks to everyone who helped us make 2025.4.0 release possible!

@Alexsandruss, @Alexandr-Solovev, @Vika-F, @david-cortes-intel, @icfaust, @napetrov, @maria-Petrova, @homksei, @ahuber21, @ethanglaser, @razdoburdin, @avolkov-intel

Full Changelog: 2025.2.0...2025.4.0

oneDAL 2025.2.0

oneDAL is happy to introduce 2025.2.0 release!

🚨 What's New

• Introduced new oneDAL functionality:
  • Implemented GPU version of spectral decomposition as linear regression fallback
  • Added ability of VTune profiling

🪲 Bug Fixes

• Fixed incorrect reduction when there's fewer bins than jobs
• Fixed rule of three violations in AlgorithmContainer classes
• Fixed additional array allocation in host to device transfer
• Fixed min, max and sum_squares computations in low order moments
• Fixed incorrect mapping of thread IDs

Acknowledgements

Thanks to everyone who helped us make 2025.2.0 release possible!

@Alexsandruss, @Alexandr-Solovev, @Vika-F, @david-cortes-intel, @icfaust, @napetrov, @maria-Petrova, @homksei, @ahuber21, @ethanglaser, @razdoburdin, @avolkov-intel

Full Changelog: 2025.1.0...2025.2.0

oneDAL is happy to introduce 2025.1.0 release!

🚩 Removals and ABI Compatibility

• Removed libonedal_sycl.a from the distribution

🚨 What's New

• Introduced new oneDAL functionality:
  • Added OpenRNG Backend
  • Added xcp reference implementation
• Improved oneDAL performance for the following algorithms:
  • SVE optimised float WSSJ kernel

🪲 Bug Fixes

• Addressed missing dependencies to compute probabilities function
• Added missing headers for finiteness_checker algorithm
• Fixed broken links in the documentation

Acknowledgements

Thanks to everyone who helped us make 2025.1.0 release possible!

@Alexsandruss, @Alexandr-Solovev, @Vika-F, @david-cortes-intel, @icfaust, @napetrov, @maria-Petrova, @homksei, @ahuber21, @ethanglaser, @samir-nasibli, @olegkkruglov, @razdoburdin, @avolkov-intel, @md-shafiul-alam

Full Changelog: 2025.0.0...2025.1.0