bfortuner / ml-study

ML Learning Sabbatical Study Materials

ML Study

Notebooks and programs for machine learning study

Setup

1. Install Anaconda
2. Install python libraries

pip install bcolz
pip install theano
pip install keras

Tools

Tutorials and cheat sheets for libraries and tools

Pandas

• Awesome Guide
• My Cheatsheet
• Quickstart
• Plotting

Numpy

• Python Data Science (Numpy, Matplotlib, Pandas)

Scikit-learn

Data School ML Basics with Scikit

Keras

• Setup

Tensorflow

Tutorial on Tensorflow

Matplotlib

• 3D Plots
• Colormaps

Tmux

• Wiki on Tmux

Resources

• Fast.ai Wiki
• Python ML Mini-Course
• First Neural Network with Keras

ml-study's Issues

TypeError: range() does not take keyword arguments

I was trying to replicate the results in the notebook and I received the above TypeError. Using py3. Any suggestions regarding that ?

visualize_runtimes(multithreading(download, urls, 1), "Single Thread")

Here is the complete traceback: -

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-13-a593e86a1f96> in <module>()
----> 1 visualize_runtimes(multithreading(download, urls, 1), "Single Thread")

<ipython-input-1-a949dc10fbff> in visualize_runtimes(results, title)
      1 def visualize_runtimes(results, title):
      2     start, stop = np.array(results).T
----> 3     plt.barh(range(len(start), stop-start, left = start))
      4     plt.grid(axis='x')
      5     plt.ylabel('Tasks')

TypeError: range() does not take keyword arguments

Question out of curiosity about multi-threading/multi-processing

Dear Mr. Fortuner,

I am Matthias, a researcher from Ghent University in Belgium and I had a small question about some of your code.
When I was programming in Python last week, I found your article 'Intro to Threads and Processes in Python'(1) on the internet. First of all, thanks a lot for sharing that, I have to say that it's very clear and interesting and I'll probably use the code for some research in the future.

But I had a small question about it. When looking at the graphs, I would love to have the threads visualized by a single color so that you can easily see which tasks were executed by which thread.

Looking a bit more into how to apply this to your code, this could probably be done by adding an extra return variable to the function 'download' that says: a = str(threading.current_thread().ident). By having the id's of each thread, you could then assign a color to each different id-number and use this list of colors as input for the visualization.
So what I was wondering is how to add an extra return variable to the 'download'-function. Now it says 'return start,stop' and I would like to have 'return start, stop, a' for example, so that the 'a' can return the id's of the threads.

But when editing the code to include this extra variable, I got nothing but errors in the parts that use the return values of the function 'download' and therefore I was kind of hoping that you maybe could give me some clues to make this happen since you know the code the best and know where you need to add what to have no conflicts.
If it finally works I can push it even to the directory as an extra feature if you want.

Kind regards,
Matthias Van Hove

(1) https://medium.com/@bfortuner/python-multithreading-vs-multiprocessing-73072ce5600b

Numpy addition multiprocessing only ran on 1 process in Multitasking notebook

Incredible work here! I noticed one small issue: in the multitasking_python.ipynb notebook in the Numpy Addition section under the Multiprocessing subsection, the notebook code does not run the addition on multiple processes. The code is the exact same for all three runs:

visualize_runtimes(multiprocessing(addition, [i for i in range(N)], 1),MULTIPROCESSING_TITLE)

I think it accidentally runs the addition on 1 process 3 times and the results are the same.