Brian Womack

Sponsored by

Intel

Creating data labels to adapt to contextual change

Date: This event took place live on February 20 2018

Presented by: Brian Womack

Duration: Approximately 60 minutes.

Cost: Free

Questions? Please send email to




Watch the webcast in its original format.

Sign in to Register

Description:

We’re experiencing a second wave of machine intelligence. Today, traditional artificial intelligence (AI) and machine learning (ML) largely depend upon data scientists to formulate labels associated with feature vectors that take into account model attributes such as entity, action, or relationship. This process can result in implementation difficulties and errors over time.

A key challenge is that a given feature vector could be a blended version of multiple labels, and many algorithms force a choice of a single label. A feature vector may also need a percentage of a given class label (e.g. 60% friendly and 35% extroverted). There is also the chance that the data scientist did not choose the best set of class labels for the data to reflect contextual change, motivating the need to retrain the algorithm models.

In this webcast, we’ll discuss:

  • Challenges of labeling data for next-generation ML applications, and suggestions for mitigating them
  • Instance-based learning methods from cognitive science and how they can create associations that can be grouped into a dynamic set of labels from which models can re-learn
  • Example uses cases to illustrate the challenges for the community to address

About Dr. Brian Womack, Senior Director at Intel Saffron

Brian is a veteran of defense intelligence and operations communities with a combination of data science, analytics algorithm development, robust signal processing, and software engineering experience. He values staying involved with the implementation details of adaptive computing technologies to advance the state of the art of human machine intelligence (HMI).

His research and development focus is on the third wave of machine intelligence (3MI) to implement Saffron's complementary learning vision; which increases system autonomy by creating a true partnership of human and machine. By integrating both traditional statistical learning methods from the second wave of machine intelligence (2MI) and instance learning methods from cognitive memory-based computing, it is possible to make high impact decisions faster with more relevant data.

Dr. Womack received his Ph.D. from Duke in robust signal processing and his M.S. from Texas A&M in adaptive control systems. Brian enjoys martial arts, SCUBA, camping, hiking, canoeing, archery, and volunteering.