machine learning

Not Just Another Pretty Face

Take a look at the photographs below and see if you can detect what is unusual about these people. Look carefully!

Give up?

The answer is that these are not real people — they are photographs generated by analytics.

Using a type of unsupervised machine learning called a generative adversarial network (GAN), Phillip Wang developed the website which dynamically generates a new human face every time the page is refreshed. His model was trained on 70,000 human photographs, and demonstrates how effectively machines can not only process image data, but can be trained to generate new data based on observable patterns.

One attribute of GANs is that it puts two analytical models in competition: one model generates the fake photos, and the other model tries to distinguish real from fake. These exact same techniques are already being used in healthcare to develop analytical approaches to detecting pathologies in clinical imaging!

CXO Insights Article: The Coming Era of High Performance Medicine

CIO Review recently published an article I wrote regarding how the health care industry can increase the value from their information technology assets. The article, called The Coming Era of High Performance Medicine, focusing on the intersection of enterprise architecture and data sciences.