Google ScholarHighlighted Work
- Model-based RL based on Dreamer.
Evolving Spatially Embedded Recurrent Spiking Neural Networks for Control Tasks
ICANN 2025 (Best Student Paper Award)
A Vasilache, J Scholz, Y Sandamirskaya, J Becker
- Spatial Embedding to optimize RNN topology for control tasks.
- SOTA performance on simple tasks using 1.46% of the weights
- 66% SOTA performance on complex tasks using 0.06% of the weights (200 vs. 350.000)
- Up to 5000x less estimated energy
Training Neural Networks by Optimizing Neuron Positions
LIVING MACHINES 2025
L Erb, T Boccato, A Vasilache, J Becker, N Toschi
- Spatial Embedding to reduce trainable parameters from O(N2) to O(N)
- Neuron positions are optimized via backprop
Spiking Neural Networks for Low-Power Vibration-Based Predictive Maintenance
ICONS 2025
A Vasilache, S Nitzsche, C Kneidl, M Tekneyan, M Neher, J Becker
- The benefits of Hardware-Software co-design on Neural Network energy consumption.
- Up to 4000x less estimated energy
A PyTorch-Compatible Spike Encoding Framework for Energy-Efficient Neuromorphic Applications
ICONS 2025 (Best Paper Award)
A Vasilache, J Scholz, V Schilling, S Nitzsche, F Kaelber, J Korsch, J Becker
- Open-source framework for converting numerical data into sparse signals.
Realtime-Capable Hybrid Spiking Neural Networks for Neural Decoding of Cortical Activity
NICE 2025
J Krausse*, A Vasilache*, K Knoblock, J Becker
- Realtime implementation of the neural decoding model from the previous publication.
- SOTA Performance on the Primate Reaching Dataset
Sleep Stage and Apnea Classification from Single-Lead ECG Using Artificial and Spiking Neural Networks
IECBES 2024 (Best Paper Award)
G Biri*, A Vasilache*, T Hu, M Themistocli, S Nitzsche, J Juhl, C Erler, S Fuhrhop, W Stork, J Becker
- Deep learning neuromorphic model for Sleep Stage Detection.
- Comparison with traditional Deep Learning Model.
- New SOTA Performance on Sleep Stage Detection.
Hybrid Spiking Neural Networks for Low-Power Intra-Cortical Brain-Machine Interfaces
BioCAS 2024 (Winner of 2nd Place in the Grand Challenge)
A Vasilache*, J Krausse*, K Knoblock, J Becker
- Deep learning neuromorphic model for neural decoding of primate motor commands.
- Model optimization for resource constrained environments.
Low-Power Vibration-Based Predictive Maintenance for Industry 4.0 using Neural Networks: A Survey
ITEM Workshop 2024
A Vasilache, S Nitzsche, D Floegel, T Schuermann, S von Dosky, T Bierweiler, M Mußler, F Kaelber, S Hohmann, J Becker
- Survey on low-power neural networks for predictive maintenance.
Awards and Achievements
- Finished PhD in 3 years at the age of 26.
- In 1 year of publications: 7 first-author, 3 best paper awards and 2nd place winner in the Grand Challenge on Neural Decoding at BioCAS 2024.
- Best New Neuromorph (BNN) Nominee at the Telluride Neuromorphic Workshop.
Short Biography
Other Interests
- Guitar
- Drawing
- Photography
- Reading
- Fitness
- Language Learning: Romanian (Native), English (C2), German (C1), Japanese (N4)