Laura Kriener

Jason Yik, Korneel Van den Berghe, Douwe den Blanken et al. · eth-zurich

Neuromorphic computing shows promise for advancing computing efficiency and capabilities of AI applications using brain-inspired principles. However, the neuromorphic research field currently lacks standardized benchmarks, making it difficult to accurately measure technological advancements, compare performance with conventional methods, and identify promising future research directions. Prior neuromorphic computing benchmark efforts have not seen widespread adoption due to a lack of inclusive, actionable, and iterative benchmark design and guidelines. To address these shortcomings, we present NeuroBench: a benchmark framework for neuromorphic computing algorithms and systems. NeuroBench is a collaboratively-designed effort from an open community of researchers across industry and academia, aiming to provide a representative structure for standardizing the evaluation of neuromorphic approaches. The NeuroBench framework introduces a common set of tools and systematic methodology for inclusive benchmark measurement, delivering an objective reference framework for quantifying neuromorphic approaches in both hardware-independent (algorithm track) and hardware-dependent (system track) settings. In this article, we outline tasks and guidelines for benchmarks across multiple application domains, and present initial performance baselines across neuromorphic and conventional approaches for both benchmark tracks. NeuroBench is intended to continually expand its benchmarks and features to foster and track the progress made by the research community.

Kevin Max, Laura Kriener, Garibaldi Pineda García et al.

Models of sensory processing and learning in the cortex need to efficiently assign credit to synapses in all areas. In deep learning, a known solution is error backpropagation, which however requires biologically implausible weight transport from feed-forward to feedback paths. We introduce Phaseless Alignment Learning (PAL), a bio-plausible method to learn efficient feedback weights in layered cortical hierarchies. This is achieved by exploiting the noise naturally found in biophysical systems as an additional carrier of information. In our dynamical system, all weights are learned simultaneously with always-on plasticity and using only information locally available to the synapses. Our method is completely phase-free (no forward and backward passes or phased learning) and allows for efficient error propagation across multi-layer cortical hierarchies, while maintaining biologically plausible signal transport and learning. Our method is applicable to a wide class of models and improves on previously known biologically plausible ways of credit assignment: compared to random synaptic feedback, it can solve complex tasks with less neurons and learn more useful latent representations. We demonstrate this on various classification tasks using a cortical microcircuit model with prospective coding.

Filippo Moro, Pau Vilimelis Aceituno, Laura Kriener et al.

Spiking Neural Networks (SNNs) have the potential for rich spatio-temporal signal processing thanks to exploiting both spatial and temporal parameters. The temporal dynamics such as time constants of the synapses and neurons and delays have been recently shown to have computational benefits that help reduce the overall number of parameters required in the network and increase the accuracy of the SNNs in solving temporal tasks. Optimizing such temporal parameters, for example, through gradient descent, gives rise to a temporal architecture for different problems. As has been shown in machine learning, to reduce the cost of optimization, architectural biases can be applied, in this case in the temporal domain. Such inductive biases in temporal parameters have been found in neuroscience studies, highlighting a hierarchy of temporal structure and input representation in different layers of the cortex. Motivated by this, we propose to impose a hierarchy of temporal representation in the hidden layers of SNNs, highlighting that such an inductive bias improves their performance. We demonstrate the positive effects of temporal hierarchy in the time constants of feed-forward SNNs applied to temporal tasks (Multi-Time-Scale XOR and Keyword Spotting, with a benefit of up to 4.1% in classification accuracy). Moreover, we show that such architectural biases, i.e. hierarchy of time constants, naturally emerge when optimizing the time constants through gradient descent, initialized as homogeneous values. We further pursue this proposal in temporal convolutional SNNs, by introducing the hierarchical bias in the size and dilation of temporal kernels, giving rise to competitive results in popular temporal spike-based datasets.

Karthik Charan Raghunathan, Christian Metzner, Laura Kriener et al.

In a continual learning setting, we require a model to be plastic enough to learn a new task and stable enough to not disturb previously learned capabilities. We argue that this dilemma has an architectural root. A finite network has limited representational and plastic resources, yet the required capacity depends on properties of the future task stream that are unknown: how many tasks will be encountered, and how much they overlap in feature space. Regularization-based methods preserve past knowledge within fixed-capacity architectures and therefore implicitly rely on an oracle architecture sized for this unknown future. When tasks are only weakly related, fixed architectures progressively run out of plastic resources; when tasks are few or strongly overlapping, models are often over-provisioned. Inspired by neurogenesis in biology, we propose NORACL to address the stability-plasticity dilemma by tackling the oracle architecture problem through neuronal growth. Starting from a compact network, NORACL grows only when needed by monitoring two complementary signals for representational and plasticity saturation. We evaluate NORACL against oracle-sized static baselines across varying task counts and geometries. Across all settings, NORACL achieves final average accuracies that are better than or on par with oracle-provisioned static baselines while using fewer parameters. Additionally, NORACL yields architectures with interpretable growth, i.e. dissimilar tasks predominantly expand feature-extraction layers, whereas tasks which rely on common features shift growth toward later feature-combination layers. Our analysis further explains why fixed-capacity networks lose plasticity as tasks accumulate, whereas NORACL creates fresh capacity for new tasks through growth. Together, these results show that adaptive neurogenesis pushes the stability-plasticity Pareto frontier of continual learning.

Benjamin Ellenberger, Paul Haider, Jakob Jordan et al.

How physical networks of neurons, bound by spatio-temporal locality constraints, can perform efficient credit assignment, remains, to a large extent, an open question. In machine learning, the answer is almost universally given by the error backpropagation algorithm, through both space and time. However, this algorithm is well-known to rely on biologically implausible assumptions, in particular with respect to spatio-temporal (non-)locality. Alternative forward-propagation models such as real-time recurrent learning only partially solve the locality problem, but only at the cost of scaling, due to prohibitive storage requirements. We introduce Generalized Latent Equilibrium (GLE), a computational framework for fully local spatio-temporal credit assignment in physical, dynamical networks of neurons. We start by defining an energy based on neuron-local mismatches, from which we derive both neuronal dynamics via stationarity and parameter dynamics via gradient descent. The resulting dynamics can be interpreted as a real-time, biologically plausible approximation of backpropagation through space and time in deep cortical networks with continuous-time neuronal dynamics and continuously active, local synaptic plasticity. In particular, GLE exploits the morphology of dendritic trees to enable more complex information storage and processing in single neurons, as well as the ability of biological neurons to phase-shift their output rate with respect to their membrane potential, which is essential in both directions of information propagation. For the forward computation, it enables the mapping of time-continuous inputs to neuronal space, effectively performing a spatio-temporal convolution. For the backward computation, it permits the temporal inversion of feedback signals, which consequently approximate the adjoint variables necessary for useful parameter updates.

Julian Göltz, Jimmy Weber, Laura Kriener et al.

Spiking neural networks (SNNs) inherently rely on the timing of signals for representing and processing information. Incorporating trainable transmission delays, alongside synaptic weights, is crucial for shaping these temporal dynamics. While recent methods have shown the benefits of training delays and weights in terms of accuracy and memory efficiency, they rely on discrete time, approximate gradients, and full access to internal variables like membrane potentials. This limits their precision, efficiency, and suitability for neuromorphic hardware due to increased memory requirements and I/O bandwidth demands. To address these challenges, we propose DelGrad, an analytical, event-based method to compute exact loss gradients for both synaptic weights and delays. The inclusion of delays in the training process emerges naturally within our proposed formalism, enriching the model's search space with a temporal dimension. Moreover, DelGrad, grounded purely in spike timing, eliminates the need to track additional variables such as membrane potentials. To showcase this key advantage, we demonstrate the functionality and benefits of DelGrad on the BrainScaleS-2 neuromorphic platform, by training SNNs in a chip-in-the-loop fashion. For the first time, we experimentally demonstrate the memory efficiency and accuracy benefits of adding delays to SNNs on noisy mixed-signal hardware. Additionally, these experiments also reveal the potential of delays for stabilizing networks against noise. DelGrad opens a new way for training SNNs with delays on neuromorphic hardware, which results in fewer required parameters, higher accuracy and ease of hardware training.

Leo Zhao, Tristan Torchet, Melika Payvand et al.

State-space models (SSMs) have recently gained attention in deep learning for their ability to efficiently model long-range dependencies, making them promising candidates for edge-AI applications. In this paper, we analyze the effects of quantization on small-scale SSMs with a focus on reducing memory and computational costs while maintaining task performance. Using the S4D architecture, we first investigate post-training quantization (PTQ) and show that the state matrix A and internal state x are particularly sensitive to quantization. Furthermore, we analyze the impact of different quantization techniques applied to the parameters and activations in the S4D architecture. To address the observed performance drop after Post-training Quantization (PTQ), we apply Quantization-aware Training (QAT), significantly improving performance from 40% (PTQ) to 96% on the sequential MNIST benchmark at 8-bit precision. We further demonstrate the potential of QAT in enabling sub-8-bit precisions and evaluate different parameterization schemes for QAT stability. Additionally, we propose a heterogeneous quantization strategy that assigns different precision levels to model components, reducing the overall memory footprint by a factor of 6x without sacrificing performance. Our results provide actionable insights for deploying quantized SSMs in resource-constrained environments.

Tristan Torchet, Christian Metzner, Laura Kriener et al.

Edge devices for temporal processing demand models that capture both short- and long- range dynamics under tight memory constraints. While Transformers excel at sequence modeling, their quadratic memory scaling with sequence length makes them impractical for such settings. Recurrent Neural Networks (RNNs) offer constant memory but train sequentially, and Temporal Convolutional Networks (TCNs), though efficient, scale memory with kernel size. To address this, we propose mGRADE (mininally Gated Recurrent Architecture with Delay Embedding), a hybrid-memory system that integrates a temporal 1D-convolution with learnable spacings followed by a minimal gated recurrent unit (minGRU). This design allows the convolutional layer to realize a flexible delay embedding that captures rapid temporal variations, while the recurrent module efficiently maintains global context with minimal memory overhead. We validate our approach on two synthetic tasks, demonstrating that mGRADE effectively separates and preserves multi-scale temporal features. Furthermore, on challenging pixel-by-pixel image classification benchmarks, mGRADE consistently outperforms both pure convolutional and pure recurrent counterparts using approximately 20% less memory footprint, highlighting its suitability for memory-constrained temporal processing at the edge. This highlights mGRADE's promise as an efficient solution for memory-constrained multi-scale temporal processing at the edge.

Sebastian Billaudelle, Laura Kriener, Filippo Moro et al.

Recurrent neural networks (RNNs) have been a long-standing candidate for processing of temporal sequence data, especially in memory-constrained systems that one may find in embedded edge computing environments. Recent advances in training paradigms have now inspired new generations of efficient RNNs. We introduce a streamlined and hardware-compatible architecture based on minimal gated recurrent units (GRUs), and an accompanying efficient mixed-signal hardware implementation of the model. The proposed design leverages switched-capacitor circuits not only for in-memory computation (IMC), but also for the gated state updates. The mixed-signal cores rely solely on commodity circuits consisting of metal capacitors, transmission gates, and a clocked comparator, thus greatly facilitating scaling and transfer to other technology nodes. We benchmark the performance of our architecture on time series data, introducing all constraints required for a direct mapping to the hardware system. The direct compatibility is verified in mixed-signal simulations, reproducing data recorded from the software-only network model.

Laura Kriener, Kristin Völk, Ben von Hünerbein et al.

Behavior can be described as a temporal sequence of actions driven by neural activity. To learn complex sequential patterns in neural networks, memories of past activities need to persist on significantly longer timescales than the relaxation times of single-neuron activity. While recurrent networks can produce such long transients, training these networks is a challenge. Learning via error propagation confers models such as FORCE, RTRL or BPTT a significant functional advantage, but at the expense of biological plausibility. While reservoir computing circumvents this issue by learning only the readout weights, it does not scale well with problem complexity. We propose that two prominent structural features of cortical networks can alleviate these issues: the presence of a certain network scaffold at the onset of learning and the existence of dendritic compartments for enhancing neuronal information storage and computation. Our resulting model for Efficient Learning of Sequences (ELiSe) builds on these features to acquire and replay complex non-Markovian spatio-temporal patterns using only local, always-on and phase-free synaptic plasticity. We showcase the capabilities of ELiSe in a mock-up of birdsong learning, and demonstrate its flexibility with respect to parametrization, as well as its robustness to external disturbances.

Paul Haider, Benjamin Ellenberger, Laura Kriener et al.

The response time of physical computational elements is finite, and neurons are no exception. In hierarchical models of cortical networks each layer thus introduces a response lag. This inherent property of physical dynamical systems results in delayed processing of stimuli and causes a timing mismatch between network output and instructive signals, thus afflicting not only inference, but also learning. We introduce Latent Equilibrium, a new framework for inference and learning in networks of slow components which avoids these issues by harnessing the ability of biological neurons to phase-advance their output with respect to their membrane potential. This principle enables quasi-instantaneous inference independent of network depth and avoids the need for phased plasticity or computationally expensive network relaxation phases. We jointly derive disentangled neuron and synapse dynamics from a prospective energy function that depends on a network's generalized position and momentum. The resulting model can be interpreted as a biologically plausible approximation of error backpropagation in deep cortical networks with continuous-time, leaky neuronal dynamics and continuously active, local plasticity. We demonstrate successful learning of standard benchmark datasets, achieving competitive performance using both fully-connected and convolutional architectures, and show how our principle can be applied to detailed models of cortical microcircuitry. Furthermore, we study the robustness of our model to spatio-temporal substrate imperfections to demonstrate its feasibility for physical realization, be it in vivo or in silico.

Laura Kriener, Julian Göltz, Mihai A. Petrovici

The Yin-Yang dataset was developed for research on biologically plausible error backpropagation and deep learning in spiking neural networks. It serves as an alternative to classic deep learning datasets, especially in early-stage prototyping scenarios for both network models and hardware platforms, for which it provides several advantages. First, it is smaller and therefore faster to learn, thereby being better suited for small-scale exploratory studies in both software simulations and hardware prototypes. Second, it exhibits a very clear gap between the accuracies achievable using shallow as compared to deep neural networks. Third, it is easily transferable between spatial and temporal input domains, making it interesting for different types of classification scenarios.

Julian Göltz, Laura Kriener, Andreas Baumbach et al.

For a biological agent operating under environmental pressure, energy consumption and reaction times are of critical importance. Similarly, engineered systems are optimized for short time-to-solution and low energy-to-solution characteristics. At the level of neuronal implementation, this implies achieving the desired results with as few and as early spikes as possible. With time-to-first-spike coding both of these goals are inherently emerging features of learning. Here, we describe a rigorous derivation of a learning rule for such first-spike times in networks of leaky integrate-and-fire neurons, relying solely on input and output spike times, and show how this mechanism can implement error backpropagation in hierarchical spiking networks. Furthermore, we emulate our framework on the BrainScaleS-2 neuromorphic system and demonstrate its capability of harnessing the system's speed and energy characteristics. Finally, we examine how our approach generalizes to other neuromorphic platforms by studying how its performance is affected by typical distortive effects induced by neuromorphic substrates.

Johannes Schemmel, Laura Kriener, Paul Müller et al.

This paper presents an extension of the BrainScaleS accelerated analog neuromorphic hardware model. The scalable neuromorphic architecture is extended by the support for multi-compartment models and non-linear dendrites. These features are part of a \SI{65}{\nano\meter} prototype ASIC. It allows to emulate different spike types observed in cortical pyramidal neurons: NMDA plateau potentials, calcium and sodium spikes. By replicating some of the structures of these cells, they can be configured to perform coincidence detection within a single neuron. Built-in plasticity mechanisms can modify not only the synaptic weights, but also the dendritic synaptic composition to efficiently train large multi-compartment neurons. Transistor-level simulations demonstrate the functionality of the analog implementation and illustrate analogies to biological measurements.