astroml / sklearn_tutorial

Tutorial for astronomy data processing with scikit-learn

Machine Learning for Astronomical Data Analysis

Note: this content is extremely out-of-date, and I would not recommend using it

If you would like a more up-to-date machine learning tutorial that grew from this content, I'd recommend the [Python Data Science Handbook](http://github.com/jakevdp/PythonDataScienceHandbook).

sklearn_tutorial's Issues

numpy.random.shuffle not work in Exercise 1

I follow the guide here. After

import os
DATA_HOME = os.path.abspath('../../../github/sklearn_tutorial/doc/data/sdss_colors')
import numpy as np
train_data = np.load(os.path.join(DATA_HOME,
                                  'sdssdr6_colors_class_train.npy'))
test_data = np.load(os.path.join(DATA_HOME,
                                 'sdssdr6_colors_class.200000.npy'))
print sum(train_data['redshift'] < 0.01)

I got 430827 row with redshift = 0. The problem is when I shuffle train_data, it fails!

nseed = 0
np.random.seed(nseed)
np.random.shuffle(train_data)
print sum(train_data['redshift'] < 0.01)

Now the result is just 288442! I change nseed to some else values, e.g. nseed = 1000000, it still doesn't work. The data changes after shuffle. Indeed, after repeat shuffle for a while, I got the data repeat itself. I mean that train_data[0] == train_data[1] == .... I believe that the shuffle function not work right. Do I miss something? I realize this after I follow the guide and scatter the train_data, and got all things are Quasars! (no Stars?).

Small bug in notebook 06_learning_curves.ipynb

In the file doc/notebooks/06_learning_curves.ipynb, I believe there is a small mistake in the code box directly under the section "Cross-validation and Testing". The data that is supposed to be used for testing is defined by:

xtest = x[Ntrain:-Ntest]
ytest = y[Ntrain:-Ntest]

Making it the same data that is used for the Cross-Validation set. I think the above code should be replaced with:

xtest = x[-Ntest:]
ytest = y[-Ntest:]

I hope this helps and thanks for putting up this great tutorial. Your book has also been really helpful

Not able to download data sets

Hello ,
I am trying to follow the scikit-learn tutorial on your website . I've setup the software as you mentioned . But i am unable to download the data sets using this command

python fetch_data.py

It says the following and stays like that until i cancel the ongoing query

downloading data from http://www.astro.washington.edu/users/vanderplas/pydata/sdssdr6_colors_class.200000.dat

What should i do to download the data sets ?

Thank you

chaithuzz2

All links to fetch data are broken

Apparently, this URL used in all fetch* scripts doesn't have the data anymore:

http://www.astro.washington.edu/users/vanderplas/pydata/

Is there a mirror to these datasets?

Broken Link on Tutorial Home Page

There are two links to Scikit-learn's home page opening paragraphs:

These learning tasks are enabled by the tools available in the open-source package scikit-learn.

scikit-learn is a Python module integrating classic machine learning algorithms in the tightly-knit world of scientific Python packages (numpy, scipy, matplotlib).

These two are broken and point to http://www.scikit-learn.org/ instead of http://scikit-learn.org/stable/

Small thing but thought I'd mention it