We are very pleased to have won three outstanding speakers for the keynote lectures.
What we remember from an episode:
Memory as reactivation, transformation, and selection
Experiences are stored in the brain via modifications of synaptic connections, changing the neural representations of specific events. Network-level signatures of these representations – “engram patterns” – can be extracted from patterns of EEG oscillations and fMRI BOLD activity. In the first part of my talk, I will show how reoccurrence of engram patterns supports diverse memory functions from short-term memory maintenance to long-term memory retrieval and consolidation: memory as reactivation. However, it is commonly assumed that memory is not a veridical reproduction of past experiences but involves substantial transformations. In the second part, I will describe a taxonomy of memory transformation processes and discuss some conceptual problems of a generative view on memory: memory as transformation. I will then describe a novel view which assumes that engram patterns consist of multiple representational formats which can be selectively activated during memory processes and quantitatively described via deep neural networks. Some initial evidence for this view of memory as selection is presented, together with ideas for future research.
Optimizing psychological treatments:
from mechanisms to predictions to clinical utility
Humboldt-Universität zu Berlin
Although psychological treatments including cognitive-behavioral therapy (CBT) work in principle to improve mental health, it appears not to work equally well for everyone: recent evidence shows that nearly every second patient suffering from an anxiety disorder fails to benefit in a clinically meaningful way – with severe consequences for patients and increasing costs for societies. Precision mental health aims to identify patients at risk for non-response already prior to treatment initialization and to improve treatments based on an in-depth mechanistic understanding. During the first part of this lecture, I will give insights into putative mechanisms that may predispose patients not to benefit from CBT, focusing on emotion regulation and its neurobiological underpinnings. In the second part, we will shift to the field of predictive analytics and its application to the prediction of treatment outcome (theranostics). Third, initial evidence and first ideas (including ethical considerations) how to implement such a theranostic pipeline into clinical utility will be discussed.
Towards a culture of computational reproducibility
Ensuring that the results of data analysis are both valid and reproducible is a fundamental responsibility of every computational scientist, but both are increasingly difficult in the context of complex analysis workflows and big data. Building off of ideas from software engineering, I will argue that we need to embrace a culture of computational reproducibility. I will outline a set of values that motivate this work and principles that guide the work, and then focus on a set of practices that can help improve reproducibility in computational science. I will conclude by addressing some potential concerns about the impacts of this cultural shift.