Machine Learning 101: How to get started with Convolutional Neural Networks

*
Accepted Session
Short Form
Intermediate
Scheduled: Thursday, June 23, 2016 from 2:30 – 3:15pm in B204

Excerpt

Machine learning and especially convolutional neural networks are on the rise. With the sheer limitless amount of data and cheap computation power, neural networks can now solve problems which have been fairly complex in the past. Cole and Hannes will demonstrate how you implement a convolutional neural network with a few lines of Python code to classify images, recognize voices or understand texts.

Description

With the fast growing amount of data, machine learning becomes more important to deal with the large amounts of data. Whether you want to classify millions of images, recognize voices or understand texts, machine learning provides you the tools to tackle all of the problems. The introduction to Machine Learning 101 will introduce one of the most prominent algorithm, called convolutional neural networks. The talk will provide you with the minimum of theory and focus on solving data challenges with machine learning.

Outline
The presentation will provide a brief introduction to neural networks, and explain convolutional neural networks. Afterwards, Cole and Hannes will introduce state-of-the-art Python libraries like Theano, Lasagne and nolearn to program convolutional neural networks. Based on the libraries, the presentation will provide examples on how to classify large amounts of images. To conclude the presentation, the talk will explain how to setup a training environment and how to advantage of Amazon’s GPUs.

Open source machine learning
The presentation will introduce the state of the art open source libraries for the use of machine learning. The presentation will focus on the Theano library, and introduce the Python wrapper packages Lasagne and nolearn.

Pre-requisites
No prior knowledge of machine learning is required. Knowledge of Python will be helpful, but isn’t required.

Tags

machine learning, neural networks, data science, python

Speaking experience

- Lecturers at the Hack Oregon University (machine learning class, presentation about neural networks and convolutional neural networks)
- Presenter at the PDX Data Science group (workshop about perceptrons, Jan 2016)
- Talk at PyDX about GeoData and Python (Oct 11th, 2015)
- PyCon Uruguay 2013 talk about learning Django (November 2nd, 2013)

Speakers

Leave a private comment to organizers about this proposal