Sharon Unsworth was awarded the Donders Cube for her hard work and commitment for the Kletskoppen festival. The evaluation committee noted that Sharon’s efforts bring an appealing message to the public, while involving many Donderians across centers. In recent years, Kletskoppen has proven to be very successful and has grown into a professional concept. As such, it includes a successful example of citizen science. The evaluation committee is convinced that the size and success of Kletskoppen goes beyond standard outreach activities. As Sharon was attending a conference, she was unable to accept the Donders Cube in person, but she was very honoured to receive the award: "Public engagement is an important part of what we do – especially in the current political climate – and it’s great to see this being recognized. I hope that in the not-too-distant future we can move to an academic system where we recognize and reward this kind of work structurally. Find out more about what we do at Kletskoppen and meet the rest of the team at our (recently revamped!) website: www.kletskoppenfestival.nl."
Donders Poster Session winners
Besides the Donders Cube, many scientific posters were also presented. Prizes were also awarded for the best poster presentations per theme. This year's winners were:
Danielle Houwing (MPI)
The left and right side of the human brain are somewhat different in structure and functions, and this 'asymmetry' can be altered in brain disorders. We examined whether mice have molecular or cellular asymmetries in their brains, since they might be useful models to understand more about the human condition. Through measuring the expression levels of hundreds of genes at high spatial resolution, we found subtle asymmetry in brain regions involved in hearing, learning and memory.
Marco Gandolfo (DCC)
“It may be crucial to detect the presence of other people even in challenging situations. In this study, we show that even in absence of any expectation for a stimulus to appear and while attention is engaged in a challenging task, participants could detect images of other people more often than other common objects (i.e. plants). This was a large-scale experiment ran in a Nijmegen’s local museum, demonstrating that we can improve generalizability of cognitive phenomena to naïve, large and diverse populations.”
Chinmaya Mishra (MPI)
We investigated the question - How can a robot signal understanding (“move on”) using visual feedback cues during a conversation? In a face-to-face conversation with a robot, we invited participants to interact with a robot where the robot exhibited one of the following visual behaviours during the interaction: nods only, long-blinks only, long-blinks with nods or no feedback. Results indicate that nodding and long-blinks with nods seem to be better at signalling understanding as compared to long-blinks only when a robot performs it in a human-robot interaction. These findings help advance our understanding of multimodal communication, and of visual addressee feedback and human-robot interaction in particular.
Zehra Kazmi (DCN)
For this study we were mainly interested in how amyloid-beta induces network hyperactivity during early stages of Alzheimer’s disease, and how different synapses contribute to this network failure. Interestingly, what we found was the underlying synaptic dysfunction were specifically prominent at inhibitory synapses leading to an overall reduction of inhibitory tuning in the networks. At the moment, we are investigating molecular mechanisms underlying this specific vulnerability of inhibitory synapses.
Umut Altin (DCC)
The first poster introduces FPGAI, an open-source compiler that converts ONNX-based neural networks into hardware designs for FPGAs. Unlike conventional tools, FPGAI combines both HLS and HDL components and supports full data transfer, training, and inference flows with additional NN training (Node Perturbation) algorithm option. The second poster builds on this by introducing a distributed version of the compiler. It targets large-scale models through an optimization framework that partitions networks across multiple FPGAs. For smaller models, we investigate virtualization techniques that allow multiple networks to run efficiently on a FPGA.