BHT
gers@bht-berlin.de
Felix Gers is one of the spokespersons for the APPL-FM project. He has been Professor at the Berlin University of Applied Sciences and Technology (BHT) in the Department of Computer Science and Media since 2005.
Felix Gers received his Diploma in Physics from the Leibniz University in Hannover. In his thesis, he worked on the non-linear process of “Stimulated Brillouin Scattering in Glass Fibers.” He completed his PhD at the EPFL (Lausanne, Switzerland) with Prof. Dr. Wulfram Gerstner, in cooperation with the laboratory of Prof. Dr. Jürgen Schmidhuber at the IDSIA (Lugano, Switzerland) in the field of AI systems (Recurrent Neural Networks with Long Short-Term Memory, better known in short as LSTM).
Felix Gers’ research focuses on Natural Language Processing, Neural Modeling, Deep Learning, as well as industrial and general applications of AI methods. In the APPL-FM project, he focuses on Continual Few-Shot Learning and Adapting Text Embeddings. His special interest is to transfer biological mechanisms and neural architectures inspired by biological brains to artificial neural networks and AI systems. In the project, he aims to address important shortcomings of foundation models such as catastrophic forgetting and requirements such as one- or few-shot learning.
Publications
- Synthetic data enables human-grade microtubule analysis with foundation models for segmentation
- Robust Conformal Prediction for Infrequent Classes
- “Where does it hurt?” – Dataset and Study on Physician Intent Trajectories in Doctor Patient Dialogues
- DeepICD-R1: Medical Reasoning through Hierarchical Rewards and Unsupervised Distillation
- Boosting Long-Tail Data Classification with Sparse Prototypical Networks
- Rethinking the Idiomaticity Decomposability Hypothesis: Evidence from Distributional Learning
- Job Interview Training with RAG-LLM: An Experimental Study with the Furhat Robot
- Comply: Learning Sentences with Complex Weights inspired by Fruit Fly Olfaction
- Robust Weight Imprinting: Insights from Neural Collapse and Proxy-Based Aggregation
- CliniBench: A Clinical Outcome Prediction Benchmark for Generative and Encoder-Based Language Models