Quantumorphic: Physics- and Quantum-Inspired Computing on the SpiNNaker2 Neuromorphic Compute Platform

Mentored by Christian Mayr
at Chair for Highly Parallel VLSI Systems and Neuromorphic Circuits, Technische Universität Dresden

In modern supercomputing, there is a trend towards alternative hardware approaches and alternative compute paradigms, as conventional CPUs are approaching scaling limits. One prominent example of such alternative approaches are quantum computers. These promise to tackle numerical problems such as monte carlo sampling or prime factorization, which are NP-hard for conventional CPU-based machines, through a mixture of extreme parallelism (quantum superposition) and probabilistic computing. Less prominent examples include for instance neuromorphic compute approaches such as the SpiNNaker2 system, which employs a massive amount of ARM microcontrollers (153 per chip, 5Mio in overall machine) that are embedded in a slim, low-latency communication fabric inspired from neuronal connections in the brain. It is a generally programmable computing substrate, but greatly differs from classical architectures like CPUs or GPUs, allowing for more distributed processing and simulation approaches.

Inspired by the brain, SpiNNaker2 contains accelerators for probabilistic computing, and it can utilize a large number of compute elements in an asynchronous fashion, bypassing the usual Amdahl limit of conventional supercomputers. Thus, we see SpiNNaker2 as a "quantumorphic" system, i.e. a system that, while not a quantum computer, shares certain characteristics with it. One particular focus of our group is thus on using SpiNNaker2 for physics/quantum-inspired algorithms. We have e.g. already shown better scaling than either quantum computers or conventional supercomputers for quadratic unconstrained binary optimization [1]. Another example could be stochastic spiking neurons for solving finite-element tesselations in a highly-parallel, asynchronous fashion [2], or highly parallel Monte-Carlo Sampling.

A tentative work plan could e.g. focus around semiprime factorization, where algorithms exist for classic supercomputers (number field sieve), quantum computers (Shor), probabilistic computers [3] and neuromorphic hardware [4], and where machine learning and/or Monte-Carlo-Sampling could assist to narrow the search space [5]. The workplan would then be (1) Analyze current approaches and their fit on SpiNNaker2. (2) choose a single algorithm or small subset for optimization and implementation. (3) Complete a full processing chain by adding parts of the classic approach (e.g. factorization construction from b-smooth numbers) and (4) hybridize/cascade/combine above approaches by networking with other PhDs of the team. A hybrid of these approaches could potentially show advantages on SpiNNaker2 that none of the individual approaches could achieve.

Thus, the work would be two-fold: (a) method development/refinement from existing SpiNNaker2 implementations. Methods could encompass neuromorphic/probabilistic versions of FEM, MC-Sampling, QUBO, etc. (b) using one or more of these methods on typical applications, such as the above factorization example. Entrepreneurial ambitions are a plus, as the above algorithms have commercial applications across many fields, from cryptography, supply chain and manufacturing optimization, via medical protein design to material science. Commercialization is intended via a dedicated new startup or one of the existing spinoffs of the chair.

Work Environment

You will be working at the Chair of Highly-Parallel VLSI Systems and Neuromicroelectronics (HPSN) at TU Dresden. At HPSN, you will work with the largest neuromorphic and real-time AI machine currently existing, the 5Mio core, 8 rack SpiNNaker2 machine. Research at HPSN focusses on real-time AI, as well as neuromorphic hardware and algorithms, plus miscellaneous circuit design topics.

Prerequisites

• Master’s Degree (or equivalent) in applied mathematics, physics, computer science or related fields of expertise.
• Very good programming skills (e.g. C++, Python, PyTorch)
• Excellent skills and practical experience in one or more of the following research areas is beneficial:
  • Hardware/Embedded Systems
  • Probabilistic computing
  • quantum computing at a logical/mathematical level
• Entrepreneurial ambitions are a plus, see description above
• Ability to collaborate well in an interdisciplinary environment
• Fluency in technical and non-technical English
• A high degree of independence, commitment, team spirit, and good communication skills.

Contact

Christian Mayr
Chair for Highly Parallel VLSI Systems and Neuromorphic Circuits
Technische Universität Dresden

Literature

[1] Chen, Zihao, et al. "ON-OFF neuromorphic ISING machines using Fowler-Nordheim annealers." Nature communications 16.1 (2025): 3086.
[2] Theilman, Bradley H., and James B. Aimone. "Solving Sparse Finite Element Problems on Neuromorphic Hardware." arXiv preprint arXiv:2501.10526 (2025).
[3] Żołnierczyk, Olgierd, and Michał Wroński. "Searching b-smooth numbers using quantum annealing: Applications to factorization and discrete logarithm problem." International Conference on Computational Science. Cham: Springer Nature Switzerland, 2023.
[4] Monaco, John V., and Manuel M. Vindiola. "Integer factorization with a neuromorphic sieve." 2017 IEEE International Symposium on Circuits and Systems (ISCAS). IEEE, 2017.
[5] Hittmeir, Markus. "Smooth Subsum Search A Heuristic for Practical Integer Factorization." International Journal of Foundations of Computer Science 35.08 (2024): 949-974.