Yngve Mardal Moe

``poetry new`` fails when creating a new system environment

14

- [x] I am on the [latest](https://github.com/python-poetry/poetry/releases/latest) Poetry version. - [x] I have searched the [issues](https://github.com/python-poetry/poetry/issues) of this repo and believe that this is not a duplicate. - [x] If...

Backend-specific functions

4

@MarieRoald and I have been thinking about how to implement backend-specific functions in TensorLy. We've currently settled on a decorator as a good approach. Here's a draft of the API...

Serialisation of decompositions

1

Decomposing large tensors can be time-consuming, and it would therefore be useful to have an easy-to-use interface for storing these decompositions to disc. I am happy to work on this...

ZeroDivisionError when `FileReader.get_batch` is called without `block_size`

1

I get a `ZeroDivisionError` when I try to read a batch without specifying what the `block_size` is. My HDF5 file is created with h5py and is a simple table. Here...

Poisson regression

1

There has been several requests for group lasso regularised Poisson regression. This was more difficult than first thought as the gradient of the Poisson loss is not globally Lipschitz. The...

With Travis stopping their free tier for open source, I have switched to GitHub Actions. This works well for the tests, but the coverage reporting with coveralls is not updated...

Selecting the regularisation coefficient can be difficult. It would therefore be useful for a utility that retrains many models (with warm start) and logs the regularisation coefficients as well as...

Currently, the only way to learn which datapoints are nonzero is by looking at the sparsity mask, however, it would be useful to have a list of the groups with...

Currently, the two-class logistic regression model and multi-class logistic regression model is implemented differently. This is because I was able to find a better bound for the Lipschitz coefficient with...

The `unfolding_dot_khatri_rao`-function is unneccesarily slow

2

Currently, the `unfolding_dot_khatri_rao` function is implemented by extracting the component vectors and using the `multi_mode_dot` function in `tenalg`. However, this is very slow (for all backends)! In fact, for smaller...