About me

Hi! I am a research associate at the Institute of Computational Neuroscience in Universitätsklinikum Hamburg-Eppendorf (UKE) specializing in bio-inspired learning and computation. My research focuses on studying structures, dynamics, functions, and computation in biological and artificial recurrent (neural) networks. I develop computational models for complex biological signals, systems, and phenomena. Additionally, with over 15 years of experience in developing algorithms for biomedical signal and image processing, my primary interest lies in method development for wearable data processing in neurology and psychiatry applications. 

I am particularly involved in three research and development projects: 

1) Exploring design space of reservoir computing (RC) by introducing innovative classes of bio-inspired wiring diagrams. Here, we study connectivity patterns of brain networks and their functional and computational implications. 

2) Developing machine learning and artificial intelligence methods to analyze wearable data in depression spectrum. 

3) Employing sophisticated lesioning and multi-site perturbation methods to dissect the contributions of individual components of artificial neural networks in task performance. Thus far, we've applied these methods to feedforward networks, CNNs, RNNs, and are expanding our investigations to encompass LLMs as well.

Soon I will start an exciting project "Validation of a Translatable Chronobiological Signature of Early Relapse in Bipolar Disorder (BD)" funded by a Wellcome Trust Mental Health Award. In this project, I will be developing machine learning models for relapse prediction in individuals living with BD. Moreover, I have the opportunity to revisit the complex model I previously developed for distrusted circadian rhythm in BD. 

Earlier, I spent three years of amazing postdoctoral research in MINDS research group working with professor Herbert Jaeger, one of the pioneers of reservoir computing. As a part of an EU project (NeuRAM3), our group worked on spiking reservoir computing and its implementation on neuromorphic hardwares. Specifically, within a setting at the crossroads between machine learning, theory of computing, computational neuroscience, and medical applications, I set up a machine learning benchmark in the domain of online biosignal processing, optimized a reservoir computing based neural learning algorithm, in a standard full-precision implementation (Matlab) on a digital computer and succeeded to significantly surpass the documented SoA on the chosen benchmark. For the same task, I also designed a spiking reservoir computer simulated on BRIAN. 

I received my PhD in biomedical engineering from Amirkabir University of Technology (Tehran Polytechnic), Iran. In my PhD thesis, to model the dynamics of mood swings in bipolar disorders, I have proposed a novel complex model based on the notion of competition between recurrent maps, which mathematically represent the dynamics of neural activation in excitatory (Glutamatergic) and inhibitory (GABAergic) pathways in the brain.

I am strongly interested in biomedical signal and image processing algorithms, applications, and hardware implementation. Among the conventional and machine-learning processing methods, I am more involved in research on recurrent neural networks for spatiotemporal modeling and analysis where I found it very relevant to study dynamical aspects of information processing in the brain. 

Here is my Google Scholar profile and I'm also active in LinkedIn, Research gate and Twitter.

Fatemeh Hadaeghi

Postdoctoral Fellow for Computational Neuroscience

Institute of Computational Neuroscience

Universitätsklinikum Hamburg-Eppendorf (UKE)

Martinistrasse 52 , 20246 Hamburg, Germany

Email:  f.hadaeghi@uke.de