Computational Modeling of the Development of Abstract Object Representations
We are pleased to announce that Jochen Triesch PhD, will speak at our next BRC Colloquium on May 27th Jochen is professor and scientist at the Goethe University Frankfurt and studies Artificial Intelligence, Robotics and Computational Neuroscience.
Abstract talk
What are the origins of abstract knowledge about objects? Infants and toddlers learn about objects quite differently from today’s artificial intelligence systems. To better understand these processes we develop computational models of how infants and toddlers acquire abstract object representations and the ability to recognize objects and object categories independent of viewpoint and other sources of variability. These models employ novel learning approaches for deep neural networks that exploit the temporal and/or multimodal structure of sensory information during extended interactions with objects. We harness head-mounted eye tracking data from toddlers to train these computational models with toddlers’ first person visual input, demonstrating that strong object representations can result from just minutes of such experience. In particular, we highlight the benefits of toddlers’ gaze behavior for successful learning. In a complementary approach, we consider learning in models receiving computer rendered visual inputs, where we can precisely control the visual or multimodal input statistics. Concretely, we show how additional linguistic input, even if rare and noisy, promotes the formation of abstract object categories. Furthermore, we demonstrate how our time-based learning approach can lead to the emergence of very abstract concepts such as „kitchen object“ or „garden object“. Finally, we study the role of behavior and knowledge of executed manipulation actions (e.g. how an object was turned) and demonstrate how exploiting this additional information can enrich the learned representations. Overall, we elucidate what computational principles seem to underly the emergence of abstract object representations in infants and toddlers.
Hybrid meeting
This meeting will be a hybrid meeting. We encourage everyone to join us in person at the Max Planck Institute for Psycholinguistics (room 163) or join via the zoom link below (Meeting ID: 268 487 1397; passcode: 85463).
