A5047389910- Advanced Memory and Neural Computing
- Ferroelectric and Negative Capacitance Devices
- Neural Networks and Reservoir Computing
- Neural dynamics and brain function
- CCD and CMOS Imaging Sensors
- Neuroscience and Neural Engineering
- Neural Networks and Applications
- Photoreceptor and optogenetics research
Kirchhoff (Germany)
2017-2023
Heidelberg University
2017-2023
Institute for Physics
2017-2020
Heidelberg University
2017
Significance Neuromorphic systems aim to accomplish efficient computation in electronics by mirroring neurobiological principles. Taking advantage of neuromorphic technologies requires effective learning algorithms capable instantiating high-performing neural networks, while also dealing with inevitable manufacturing variations individual components, such as memristors or analog neurons. We present a framework resulting bioinspired spiking networks high performance, low inference latency,...
Emulating spiking neural networks on analog neuromorphic hardware offers several advantages over simulating them conventional computers, particularly in terms of speed and energy consumption. However, this usually comes at the cost reduced control dynamics emulated networks. In paper, we demonstrate how iterative training a hardware-emulated network can compensate for anomalies induced by substrate. We first convert deep trained software to BrainScaleS wafer-scale system, thereby enabling an...
We present first experimental results on the novel BrainScaleS-2 neuromorphic architecture based an analog neuro-synaptic core and augmented by embedded microprocessors for complex plasticity experiment control. The high acceleration factor of 1000 compared to biological dynamics enables execution computationally expensive tasks, allowing fast emulation long-duration experiments or rapid iteration over many consecutive trials. flexibility our is demonstrated in a suite five distinct...
The massively parallel nature of biological information processing plays an important role due to its superiority in comparison human-engineered computing devices. In particular, it may hold the key overcoming von Neumann bottleneck that limits contemporary computer architectures. Physical-model neuromorphic devices seek replicate not only this inherent parallelism, but also aspects microscopic dynamics analog circuits emulating neurons and synapses. However, these machines require network...
Abstract This paper presents verification and implementation methods that have been developed for the design of BrainScaleS-2 65 nm ASICs. The 2nd generation BrainScaleS chips are mixed-signal devices with tight coupling between full-custom analog neuromorphic circuits two general purpose microprocessors (PPU) SIMD extension on-chip learning plasticity. Simulation automated analysis pre-tapeout calibration highly parameterizable neuron synapse hardware-software co-development digital logic...
Neuromorphic systems open up opportunities to enlarge the explorative space for computational research. However, it is often challenging unite efficiency and usability. This work presents software aspects of this endeavor BrainScaleS-2 system, a hybrid accelerated neuromorphic hardware architecture based on physical modeling. We introduce key Operating System: experiment workflow, API layering, design, platform operation. present use cases discuss derive requirements showcase implementation....
Despite being originally inspired by the central nervous system, artificial neural networks have diverged from their biological archetypes as they been remodeled to fit particular tasks. In this paper, we review several possibilites reverse map these architectures biologically more realistic spiking with aim of emulating them on fast, low-power neuromorphic hardware. Since many devices employ analog components, which cannot be perfectly controlled, finding ways compensate for resulting...
A fruitful approach towards neuromorphic computing is to mimic mechanisms of the brain in physical devices, which has led successful replication neuronlike dynamics and learning past. However, there remains a large set neural self-organization whose role for yet be explored. One such mechanism homeostatic plasticity, recently been proposed play key shaping network correlations. Here, we study---from statistical-physics point view---the emergent collective homeostatically regulated device...
To rapidly process temporal information at a low metabolic cost, biological neurons integrate inputs as an analog sum but communicate with spikes, binary events in time. Analog neuromorphic hardware uses the same principles to emulate spiking neural networks exceptional energy-efficiency. However, instantiating high-performing on such remains significant challenge due device mismatch and lack of efficient training algorithms. Here, we introduce general in-the-loop learning framework based...
The BrainScaleS-2 (BSS-2) Neuromorphic Computing System currently consists of multiple single-chip setups, which are connected to a compute cluster via Gigabit-Ethernet network technology. This is convenient for small experiments, where the neural networks fit into single chip. When modeling larger size, neurons have be across chip boundaries. We implement these connections BSS-2 using EXTOLL networking provides high bandwidths and low latencies, as well message rates. Here, we describe...
A unique feature of neuromorphic computing is that memory an implicit part processing through traces past information in the system's collective dynamics. The extent about inputs commonly quantified by autocorrelation time Based on experimental evidence, a potential explanation for underlying autocorrelations are close-to-critical fluctuations. Here, we show self-organized networks excitatory and inhibitory leaky integrate-and-fire neurons can originate from emergent bistability upon...
We demonstrate five experiments that are performed on the novel BrainScaleS-2 neuromorphic architecture based an analog neuro-synaptic core and augmented by embedded microprocessors for complex plasticity experiment control. The high acceleration factor of 1000 compared to biological dynamics enables execution computationally expensive tasks, allowing fast emulation long-duration or rapid iteration over many consecutive trials. were chosen emphasize different aspects system, especially...
This paper presents verification and implementation methods that have been developed for the design of BrainScaleS-2 65nm ASICs. The 2nd generation BrainScaleS chips are mixed-signal devices with tight coupling between full-custom analog neuromorphic circuits two general purpose microprocessors (PPU) SIMD extension on-chip learning plasticity. Simulation automated analysis pre-tapeout calibration highly parameterizable neuron synapse hardware-software co-development digital logic software...
Neuromorphic systems open up opportunities to enlarge the explorative space for computational research. However, it is often challenging unite efficiency and usability. This work presents software aspects of this endeavor BrainScaleS-2 system, a hybrid accelerated neuromorphic hardware architecture based on physical modeling. We introduce key Operating System: experiment workflow, API layering, design, platform operation. present use cases discuss derive requirements showcase implementation....