Visualizing the Learning in ML Algorithms
by Jaidev Deshpande (speaking)
Objective
Mostly useful for machine learning education, this talk provides an insight into the different processes of how a classifier or a regression analysis adapts itself to a random dataset. In contrast, this might also revel interesting features of the data itself.
Description
Data alone is meaningless. It can be made sense of only in terms of the operations carried out on it. Of these operations, machine learning is one of the most versatile and difficult. This talk will focus on visualizing the learning process in popular machine learning algorithms like perceptrons, backpropagation (stochastic gradient descent), and support vector machines. The machine learning algorithms used will be from the scikit-learn package.
The visualizations will be built using Chaco, Enthought's 2-D plotting library and the talk will also feature a data-explorer GUI made in Enaml, Enthought's new DSL for pythonic UIs.
While the most immediate benefit of the talk is to students of machine learning, the visualizations can also be used to explore data in general, as we shall see.
Speaker bio
I'm an electrical engineer and an amateur in machine learning and signal / image processing. Currently an intern at Enthought, Inc, working on data analysis and visualization.
2
▼
Please add more content in the description.
1
▼
Yes, I will be adding more specifics within over the next week, as I shape my fragmented work into a PyCon 'talk'.