- ScipyにはSparse matrix向けのQR分解がありません...
- Dense matrix向けはあります!が,todense()するとメモリで死んでしまいます.
- SuiteSparseを使うとできるけどpythonで使えない,
- Python wrapperのPySPQRを使う.
- Windows 10
- Microsoft Visual Studio 14.0 (SuiteSparseのコンパイルにいる)
$ jupyter notebook
...
OSError: [Errno 99] Cannot assign requested address
こういう風にいれましょう.
$ jupyter notebook --ip=127.0.0.1
Numpyで複数のndarrayを保存するとき,npzを使うのは便利だが
import numpy as np
X = np.zeros((100,100))
y = np.zeros(10)
file_path = /home/samejima/data.npz
np.savez(file_path, x = X, y=y)
When passing big arrays to joblib.Parallel in docker container, parallel processing does not start immediately.
Joblib uses the folder specified by "JOBLIB_TEMP_FOLDER" for memmap of the arrays. Without specifying, JOBLIB_TEMP_FOLDER is set to /dev/shm
, which usually has small size and is not enough for the big arrays.
https://pythonhosted.org/joblib/generated/joblib.Parallel.html
This is how to install Fujixerox C5000-d printer in ubuntu 16.04. Unfotunately, the official printer driver is too old to be compatibe with latest ubuntu.
tar Jxfv fxlinuxprint_1.1.0+ds.orig.tar.xz
import os | |
import argparse | |
import tensorflow as tf | |
# global variables | |
width = 28 | |
height = 28 | |
n_class = 10 | |
# tfrecord parser |