YSS 23YSS 23WHEN: Thursday, 9 November
WHERE: Institute of Science and Technology Austria > Moonstone Building (I24) > foyer and seminar rooms on ground floor
09:15 – 10:00 Registration, badge distribution and coffee break
10:00 – 11:00 Plenary Talk by Giovanni Volpe (Deep learning for microscopy and neurosciences)
11:00 – 11:30 Parallel session
11:30 – 11:55 Coffee break
12:00 – 13:00 Plenary Talk by Alex M. Bronstein (Human and artificial intelligence in structural biology and organic chemistry)
13:00 – 14:25 Lunch
14:30 – 15:30 Plenary Talk by Athena Akrami (Learning and exploiting sensory statistics across multiple species)
15:30 – 16:00 Parallel Session
16:00 – 16:25 Coffee break
16:30 – 17:00 Parallel Session
17:00 – 17:30 Parallel Session
17:30 – 18:30 Plenary Talk by Angelo Cangelosi (Developmental robotics for language learning, trust and theory of mind)
18:30 – 20:00 Social beers/snacks break
Deep learning for microscopy and neurosciences – Giovanni Volpe (10:00 – 11:00)
In the rapidly evolving landscape of microscopy and neurosciences, deep learning emerges as a pivotal tool for data interpretation and innovation. This presentation delves into the journey of developing DeepTrack and Braph, two open-access software tools that epitomize this integration. DeepTrack, crafted for precision in digital video microscopy, and Braph, tailored for advanced brain analysis using graph theory, both stem from extensive research at the Soft Matter Lab. The talk emphasizes the collaborative and open-access nature of these tools, underscoring their potential to democratize advanced research techniques and foster community-driven advancements in their respective fields.
Human and artificial intelligence in structural biology and organic chemistry – Alex Bronstein (12:00 – 13:00)
In this talk, I will share my personal perspective of a computer scientist developing statistical and machine-learning methods to address problems in structural biology and organic chemistry. The first part of the talk will focus on local protein structure. The fact that some amino acid chains fold alone into natively structured and fully functional proteins in solution, has led to the commonly accepted “one sequence-one structure” dogma. However, within the cell, protein chains are not formed in isolation, to fold alone once produced. Rather, they are translated from genetic coding instructions (for which many versions exist to code a single amino acid sequence) and begin to fold before the chain has fully formed through a process known as co-translational folding. The effect of coding and co-translational folding mechanisms on the final protein structure is not well understood and there are no studies showing side-by-side structural analysis of protein pairs having alternative synonymous coding. I will overview our recent large-scale computational studies exploiting the wealth of high-resolution protein structures in the Protein Data Bank to explore the association between synonymous genetic coding and local protein structure and pinpoint positions of alternate conformations in protein backbones that cannot be readily explained by the amino acid sequence or protein environment. The second part of the talk will focus on building generative models for molecular design conditioned by various properties. I will highlight a guided diffusion model operating on a carefully constructed representation space of polycyclic aromatic systems with unprecedented near 100% validity of generated molecules.
Learning and exploiting sensory statistics across multiple species – Athena Akrami (14:30 – 15:30)
A defining feature of animal intelligence is the ability to discover and update knowledge of statistical regularities in the sensory environment, in service of adaptive behaviour. This allows animals to build appropriate priors, in order to disambiguate noisy inputs, make predictions and act more efficiently. Despite decades of research in the field of human cognition and theoretical neuroscience, it is not known how such learning can be implemented in the brain. We took a cross-species approach by developing well-controlled comparative paradigms in humans, rats, and mice. We compared their performance on a 2-alternative-forced-choice (2AFC) sound categorization task, where stimulus statistics were carefully manipulated without affecting the overall weight of each category. We investigated decision-making in such distinct “statistical contexts” associated with different stimulus prior probabilities. All species optimally adapted their decisions given the underlying sound statistics. Despite the overall sensory-dependent adaptation that is similar across species, the learning speed and trial-to-trial learning updates show interesting individual variabilities. Humans learn the statistics fastest, and most of them form a generative understanding of statistics, by learning the two category distributions separately. Some humans, however, similar to most of mice and some rats, only learned the boundary between the two sound categories, without learning each category separately. Combining rats and mice can span the full range of variability observed in the human and put us now in a unique position to investigate their brains to see how they function differently, when recruiting different strategies.
Developmental robotics for language learning, trust and theory of mind – Angelo Cangelosi (17:30 – 18:30)
Growing theoretical and experimental research on action and language processing and on number learning and gestures clearly demonstrates the role of embodiment in cognition and language processing. In psychology and neuroscience, this evidence constitutes the basis of embodied cognition, also known as grounded cognition (Pezzulo et al. 2012). In robotics and AI, these studies have important implications for the design of linguistic capabilities in cognitive agents and robots for human-robot collaboration, and have led to the new interdisciplinary approach of Developmental Robotics, as part of the wider Cognitive Robotics field (Cangelosi & Schlesinger 2015; Cangelosi & Asada 2022). During the talk we will present examples of developmental robotics models and experimental results from iCub experiments on the embodiment biases in early word acquisition and grammar learning (Morse et al. 2015; Morse & Cangelosi 2017) and experiments on pointing gestures and finger counting for number learning (De La Cruz et al. 2014). We will then present a novel developmental robotics model, and experiments, on Theory of Mind and its use for autonomous trust behavior in robots (Vinanzi et al. 2019, 2021). The implications for the use of such embodied approaches for embodied cognition in AI and cognitive sciences, and for robot companion applications will also be discussed.