This talk will be a playful but serious introduction to natural language processing and image processing with (artificial) neural networks.
Collecting data is more fun when it's about myself: I'm an expert! So I program my computer to spy on me. And then I play with the data, the better to understand both ML techniques and what I've been up to. In the limit, I try not only to understand how I spend my time but also to predict how I am about to spend my time.
Consider the active window on my computer: the window with which I am currently interacting. Watching the window name creates a sequence of window titles and a great NLP playground. ("Natural language" is here a discovered irony.) Snapshotting the contents of the window in turn creates long sequences of images. To save space, I reduce the images to postage stamp size, and yet it turns out I can learn some very interesting things nonetheless. Some reasonably simple neural networks are enough to discern how I spend my time, when I work on which projects. Indeed, well beyond providing a great playground for learning about ML techniques, the project also provides a good reminder (warning?) about how much we can learn from what appear to be relatively scant data. Of course, the links in the final slides will share all the code to help you get started with projects of your own.