Motivation for the Symposium

Deep-​learning, a brain-​inspired weak form of AI, allows training large ANNs (artificial neuronal networks) that, like humans, learn real-​world tasks such as recognizing speech or objects in images. The origins of deep learning can be traced back to early neuroscience research, for example, the work by Hubel and Wiesel in the 1960s, who first described hierarchical neuronal processing of visual inputs in the mammalian neocortex.

Similar to the brain, ANNs seem to learn by interpreting and structuring the ‘training’ data provided by the external world. However, while on specific tasks such as playing video games, deep ANNs outperform humans, they are still not on par to act as general problem solvers and their learning and abstraction capabilities are still far behind of what the human brain seems to achieve effortlessly. Further, biological neuronal networks (BNNs) in the brain learn far more effectively with fewer training examples, they achieve a much higher performance in recognising complex patterns in time series data (e.g. recognizing actions in movies), they dynamically adapt and learn new tasks without losing performance and they achieve unmatched generalization performance to detect and integrate out-​of-domain data examples (data they have not been trained with).

The examples above illustrate that many of the big challenges and unknowns in the field of deep learning are mastered exceptionally well by the brain. However, it has remained a mystery how biological networks implement such remarkable learning capabilities. Recent neuroscience evidence suggests that learning in biological networks is the result of a complex interplay of diverse error feedback signaling mechanisms that act at multiple time and spatial scales, ranging from single synapses to entire networks. Untangling the parallels and differences of learning in biological and artificial neuronal networks to jointly advance both fields, is one of the main goals of this interdisciplinary symposium.