Priberam Machine Learning Lunch Seminars (T10) - 5 - "An optimization approach for structured agent-based provider/receiver tasks", Kim Baraka (IST / CMU)

Skip to first unread message

Zita Marinho

Mar 25, 2019, 6:59:16 AM3/25/19
Hello all,

Next week Kim Baraka, a PhD student from CMU/Portugal Program will present his work on "An optimization approach for structured agent-based provider/receiver tasks" on Thursday 4th April at 13:00h (room PA2- Pav. Matemática).

Due to a large number of leftovers, we kindly ask you to please register only if you intend to attend the event:

Best regards,

Zita Marinho

Priberam Labs

Priberam is hiring!
If you are interested in working with us please send your info to
Image result for priberam logoPRIBERAM SEMINARS   --  Room PA2

Priberam Machine Learning Lunch Seminar
Speaker:  Kim Baraka (CMU / IST)
Venue: IST Alameda, Room PA2 (Pavilhão de Matemática)
Date: Thursday, April 4th, 2019
Time: 13:00

Lunch will be provided


An optimization approach for structured agent-based provider/receiver tasks


This work contributes an optimization framework in the context of structured interactions between an agent playing the role of a ‘provider’ and a human ‘receiver’. Examples of provider/receiver interactions of interest include ones between occupational therapist and patient, or teacher and student. We specifically consider tasks where the provider agent needs to plan a sequence of actions, where actions have associated costs and are organized along a hierarchy with increasing probabilities of success. The goal of the provider is to achieve a success with the lowest overall cost possible. In our application domains, a success may be for instance eliciting a desired behavior or a correct response from the receiver. Based on our mathematical formulation, we present a linear-time optimal planning algorithm that generates cost-optimal sequences for given action parameters. We provide proofs for a number of properties of optimal solutions including the fact that, for appropriate parameter selection, the resulting sequences are nondecreasing according to the action hierarchy, which aligns with typical provider strategies. Finally, we instantiate our general formulation in the context of a robot-assisted therapy task for children with Autism Spectrum Disorders (ASD). In this context, we present methods for determining action parameters, namely (1) an online survey with experts for determining action costs, (2) a probabilistic model of response to robot actions based on data collected in a real interaction scenario with 10 ASD children and a humanoid robotic agent.

Short Bio:

Kim Baraka is currently a dual degree Ph.D. student at Carnegie Mellon’s Robotics Institute (Pittsburgh, PA, USA), and Instituto Superior Técnico (Lisbon, Portugal), co-advised by Manuela Veloso and Francisco Melo. He holds an M.S. in Robotics from Carnegie Mellon, and a Bachelor in Electrical and Computer Engineering from the American University of Beirut. He was a summer student at CERN, and a recipient of the IEEE Student Enterprise Award. His research interests lie at the intersection of Artificial Intelligence, Machine Learning and Human-Robot Interaction, aimed at making robots more adaptive and more transparent to humans. His doctoral thesis focuses on the use of Artificial Intelligence to enrich social interactions between robots and humans, specifically in the context of robot-assisted autism therapy. In parallel from his scientific work, he is a professionally trained contemporary dancer, performing, teaching, and creating artistic work.

More info:


Reply all
Reply to author
0 new messages