PierrickPochelu
PierrickPochelu
https://github.com/numenta/NAB/blob/master/nab/labeler.py line 424 The following code: ``` front = max(a - windowLength/2, 0) back = min(a + windowLength/2, length-1) windowLimit = [strf(data["timestamp"][front]), strf(data["timestamp"][back])] ``` Produces: `KeyError` due to the fact...
This rescale is generally useless because reduce_sum is generally equal to 1. Disabling the rescale is significantly faster (E.g, from 11 seconds to 7 seconds). Why reduced_sum(output) is always equal...
In `categorical_crossentropy`, I suspect this normalization line to be not useful and leads to 2 unexpected behaviors https://github.com/keras-team/keras/blob/b80dd12da9c0bc3f569eca3455e77762cf2ee8ef/keras/backend.py#L5540 The upper block handle logits and returns something if it is a...
Hello, It is a great initiative to bring together all timeseries frameworks in the same place. There is no mention of audio frameworks (e.g., librosa). I can complete the list...
**Description** My laptop is equipped with multi-cores. Increasing n_jobs improve computing speed of `fit` and `predict`. However, it is inefficient for improving the speed of `predict_survival_function`. **Code Sample to Reproduce...
Kernel methods may take several order magnitude more time than regressors. I propose to introduce a time threshold (t) seconds for training algorithms. When the training time of an algorithm...
Hello, I am currently modeling multiple producers using Boost.Coroutine2 and storing them in a container (std::vector). My goal is to leverage OpenMP to accelerate the calls of these coroutines on...
In `categorical_crossentropy`, I suspect this normalization line to be not useful and leads to 2 unexpected behaviors https://github.com/keras-team/keras/blob/b80dd12da9c0bc3f569eca3455e77762cf2ee8ef/keras/backend.py#L5540 The upper block handle logits and returns something if it is a...