Source code for the "Discussions on Machine Learning" article.
Small data files and datasets that we have used as part of these notebooks are stored in the
data/ folder. Larger data files are downloaded as part of their respective notebooks' code.
If you would like to interact with these notebooks, you are strongly recommended to use Binder (What is Binder?) rather than downloading this repository onto your local machine, because, aside from the fact that Binder provides an environment with all the required Python packages preinstalled, it also enables you to work with the datasets used in these notebooks - especially those which are large - without downloading them onto your local machine. Launch the Binder environment by clicking the badge above.
While using Binder, remember that all your work is lost when you close the browser tab containing the home page of your Binder notebook environment. Also, Binder has a short timeout for sessions - according to this source it is 20 minutes, though code can run for 12 hours or longer. ('Timeout' essentially means that if you leave the Binder environment inactive, that is, you do not use the notebooks, terminals or files, for more than a certain period of time, the Binder session will expire and all your work will be lost.) This is especially problematic while going through the longer notebooks, for which one sitting may not be enough. As of 2020, the Binder interface for the notebook files has three buttons to help you work around this problem, as detailed here. Basically
- Before you go for a break, for all notebooks that you have open, click on the 'Save to browser storage' button.
- After returning from your break, if you find that the Binder session has expired - if the open notebooks are unable to connect to the kernel, and on reloading the home page of the Binder environment you get a 404 error - then you must launch a new Binder environment using the above badge.
- Finally open all the notebooks that you were working on previously and click on the 'Restore from browser storage' button, to restore all of them to their state prior to the break.
Note that this will restore your notebooks, but not the other files! In other words, in notebooks that involve writing to the filesystem, in the form of downloading datasets, creating new files, or changing the content of existing files, please ensure that you are done with all such tasks before you take a break, otherwise you may have to repeat some of your work again!
These notebooks can alternatively be viewed as slideshows: for each notebook, we have a corresponding
.slides.html file inside the
slides/ folder. You can download this file and open it using your browser to view a presentation running inside your browser itself. Note that these slides are not perfect, because they have been created using the default slideshow facility of Jupyter Notebooks (What are Jupyter Notebook slides?), so there are cases in which, for example, the output of a code cell is too long due to which the full output is not visible in the corresponding slide. However, I believe that this a concise way to present the content of the notebooks.