Keisuke Fujii

I recently raised an issue of nan-aggregation in numpy/numpy#10456, but it looks a little hard to implement nan-aggregations without copy.

Thanks, @hameerabbasi. How about to enable the use of general `scipy.sparse`-like arrays (and also `np.ndarray`-like array) in sparse? I think it would have done connection to #86, when we may...

> If Numba doesn't properly support these NumPy-ish arrays, then we run into issues. Numba does not support Cupy... OK. It looks the GPU support is a little further away...

Thanks @kwgoodman . > If it is then the last line of the function (in C) is return PyFloat_FromDouble(asum); Some leading digits looks different from `np.nansum(a)` and also from `np.nansum(a.astype(float))`....

@dcherian I think the problem is how to serialize `MultiIndex` objects rather than the array itself. In your `encoded`, how can we tell the MultiIndex is `[('a', 1), ('b', 1),...

@dcherian. Now I understood. Your working examples were really nice for me to understand the idea. Thank you for this clarification. I think the use of this convention is the...

@davharris Thanks for the information. It would help! FYI, I am using the following workaround for the gradient clipping. ```Python class ClippedAdamOptimizer(tf.train.AdamOptimizer): """ Clipped version adam optimizer, where its gradient...

Closed because we discuss this issue in #33.

Hi @TomAugspurger Do you know when the test starts failing? This PR does not seem relevant.

I tried an investigation, and seems like an upstream issue. Rose an issue (there)[https://github.com/scikit-learn/scikit-learn/issues/29534]. With numpy==2.0, it seems like that `sklearn.decomposition.PCA` is unstable, sometimes giving strange values.