Main Arraymancer contributors (sorted alphabetically)
- Autograd of mean along an axis
- Add features that were missing from Numpy and Matlab, such as:
- Missing math functions and operators (contains, convolve, diff_discrete, median, sinc, roll, flatten, append, moveaxes, unique, etc.)
- Add or improve diagonal, triangular and other special matrix functions (toepliz, hankel, vander, etc.)
- Improve Complex and bool tensor support as well as mixed real-Complex ops
- Add support for combining span slices with negative steps (e.g.
t[_|-1]) - Add support for doing a masked fill from a tensor
- Remove some compilation warnings
- Error message and documentation improvements
- Miscellaneous fixes
- OpenMP
- Several performance optimizations and fixes including
- Strided iterators
- Uninitialized seq
- Shapeshifting procs
- Developing the ecosystem with arraymancer-vision and arraymancer-demos
- CSV and toSeq exports
- Tensor plotting tool
- several fixes
- Lead dev
- tanh activation
- Kmeans clustering
- Automation of MNIST download, caching and reading from compressed gzip
- IMDB dataset loader
- HDF5