The current Ebola outbreak almost defies understanding. It has struck otherwise healthy individuals with a host of terrible symptoms. As of October 2014, Ebola has infected some 9,000 people, mostly in western Africa. Half have died. No cure exists.

What does exist is data related to the disease. A good deal of it resides in open data sets — content anyone can access without concern for copyright or patent infringement. Indeed, open data holds out hope to data scientists who want to help stem transmission rates, target resources in affected countries — and even contribute to a cure.

In this episode of Inside IBM Research, IBM Fellow Saska Mojsilovic talks about the first steps she and her colleagues at IBM are taking with the Ebola Open Data Jam and a crowdsourcing project called MapGive to create “data maps” that could help public health teams contain the spread of the disease.

See our related podcast episode:
Containing Ebola With Mathematical Modeling
https://soundcloud.com/ibmresearch/containing-ebola-with-mathematical-modeling


Will Crowdsourcing And Data Stem The Ebola Outbreak

IBM Watson is revolutionizing the way we see and find patterns in unstructured data, from research papers to images. And it's revolutionizing the way people in healthcare, finance, retail, travel and services are making critical decisions.

In this episode of Inside IBM Research, Janani Mukundan, a machine learning researcher at IBM Research-Austin, talks about how a stack of Restricted Boltzmann Machines is now building off Watson to learn how to compose music. The results are expected to have an impact on the study of machine learning and natural language.

You can read the transcript and listen to Janani's complete musical pieces on the Researcher show page.
http://researcher.watson.ibm.com/researcher/view_group.php?id=6376

The Music of Machines

Almost nothing frightens us as much as mental illness. If only we had a way of knowing who was likely to succumb to it. We could anticipate a psychotic break and, optimally, deliver an effective therapy to treat or manage it.

IBM physicist Guillermo Cecchi and his colleagues in computer science and psychiatry have developed an algorithm that infers a schizophrenic break from an individual’s language patterns. In this episode of Inside IBM Research, he talks about a multidisciplinary study that will help psychiatrists anticipate serious mental illness — and help society ease human suffering.

Check back in January 2016 for a complete transcript.

Related links

Guillermo Cecchi
http://researcher.watson.ibm.com/researcher/view.php?person=us-gcecchi

Inside IBM Research
http://researcher.watson.ibm.com/researcher/view_group.php?id=5665

Why Are Computers So Accurate at Predicting Schizophrenia? [Forbes]
http://www.forbes.com/sites/michaelthomsen/2015/09/20/why-are-computers-so-accurate-at-predicting-schizophrenia/

Computers Can Predict Schizophrenia Based on How a Person Talks [The Atlantic]
http://www.theatlantic.com/technology/archive/2015/08/speech-analysis-schizophrenia-algorithm/402265/

Look Who Is Getting Into Mental Health Research [NIH]
http://www.nimh.nih.gov/about/director/2015/look-who-is-getting-into-mental-health-research.shtml

Word Analysis Proves Effective Pre-Psychosis Diagnosis [Building a Smarter Planet]
http://asmarterplanet.com/blog/2015/08/word-analysis-proves-effective-pre-psychosis-diagnosis.html



Using An Algorithm To Predict Schizophrenia

Watson Beat is a machine deep learning algorithm that learns and creates music. IBM researchers are teaching a complex neural network to understand music theory, structure (pitch, time signature, key signature), emotional intent and co-create music with a human partner. Here are some of the results.

String Quartet Amped

Stabilize Amped

Juniper Breeze Amped

Fallen Star Amped

BachPrelude1

Juniper Breeze Original

Stabilize Original Intrumental

String Quartet Original

This is the official Soundcloud account for IBM Research. It is managed by Chelsea Dubner and Chris Nay and follows IBM social media guidelines.

IBM Research

Popular tracks