Patrick Krauss
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Patrick Krauss has a total of 133 co-authors. Here are the top co-authors by number of shared publications:
Achim Schilling
Cognitive Computational Neuroscience Group, Pattern Recognition Lab, Friedrich-Alexander-University Erlangen-Nürnberg (FAU)
51 shared publicationsClaus Metzner
Cognitive Computational Neuroscience Group, Pattern Recognition Lab, Friedrich-Alexander-University Erlangen-Nürnberg (FAU)
25 shared publications- Achim Schilling
Pattern Recog. Lab Neuroscience Lab CCN Group University Hospital Erlangen FAU Erlangen-Nürnberg Erlangen Germany
24 shared publications Andreas Maier
Cognitive Computational Neuroscience Group, Pattern Recognition Lab, Friedrich-Alexander-University Erlangen-Nürnberg (FAU)
22 shared publicationsHolger Schulze
Neuroscience Lab, University Hospital Erlangen, Friedrich-Alexander University Erlangen-Nürnberg (FAU), Germany
19 shared publicationsKonstantin Tziridis
Neuroscience Lab, University Hospital Erlangen, Germany
16 shared publications
Latest Publications
- Open Access
- 23 October 2026
- Published by Frontiers Media SA in Frontiers in Complex Systems
- Vol. 4, 1960494
Convergent evolution in neural representation space: emergent order in deep belief networks
Deep Belief Networks (DBNs) learn hierarchical generative models without class supervision. We ask how this unsupervised hierarchy reorganizes data, whether it exposes latent class structure, and whether independently initialized networks discover corresponding internal representations. Across MNIST, Fashion-MNIST, and KMNIST, we follow class geometry and accessibility together with representational dimensionality, input reconstruction, and free generation. We find that global class organization generally strengthens with depth even though labels are never available during learning. Early layers frequently make class identity easier to extract, whereas deeper layers form increasingly compact representations in which related features are learned repeatedly. Global class geometry and classifier accessibility can nevertheless diverge, showing that a more ordered representation is not uniformly more informative at every class boundary. Independently trained DBNs converge to closely related depth-specific representation geometries and neuronal response features, indicating that the learned hierarchy is strongly constrained by the data and learning process. Deeper generative paths also produce broader populations of recognizable free samples while preserving less information about individual inputs. Together, these results show that unsupervised depth follows a reproducible path of representational reorganization in which class structure, accessibility, compression, reconstruction, and generation remain distinct but interacting properties.
- 23 October 2026
- Published by Frontiers Media SA in Frontiers in Complex Systems
- Vol. 4, 1960494
Open AccessConvergent evolution in neural representation space: emergent order in deep belief networks
Deep Belief Networks (DBNs) learn hierarchical generative models without class supervision. We ask how this unsupervised hierarchy reorganizes data, whether it exposes latent class structure, and whether independently initialized networks discover corresponding internal representations. Across MNIST, Fashion-MNIST, and KMNIST, we follow class geometry and accessibility together with representational dimensionality, input reconstruction, and free generation. We find that global class organization generally strengthens with depth even though labels are never available during learning. Early layers frequently make class identity easier to extract, whereas deeper layers form increasingly compact representations in which related features are learned repeatedly. Global class geometry and classifier accessibility can nevertheless diverge, showing that a more ordered representation is not uniformly more informative at every class boundary. Independently trained DBNs converge to closely related depth-specific representation geometries and neuronal response features, indicating that the learned hierarchy is strongly constrained by the data and learning process. Deeper generative paths also produce broader populations of recognizable free samples while preserving less information about individual inputs. Together, these results show that unsupervised depth follows a reproducible path of representational reorganization in which class structure, accessibility, compression, reconstruction, and generation remain distinct but interacting properties.
- Open Access
- 20 October 2026
- Published by Frontiers Media SA in Frontiers in Computational Neuroscience
- Vol. 20, 1886601
Are cortical microcircuits optimized for information flux?—A simulation-based reverse engineering study
Introduction: A sufficiently large information flux in recurrent neural networks, quantified by the mutual information between successive network states, has been proposed as a marker of rich information-processing capabilities. This raises the question of whether biological neural networks, such as cortical microcircuits, may be structurally organized to enhance information flux. Methods: To investigate this possibility, we study a simplified model of the cortical layer 5 architecture, in which a densely and strongly interconnected core population is embedded within a larger supporting network. Results: Surprisingly, we find that the embedding network exerts a pronounced flux-enhancing effect on the core dynamics. Systematic reverse-engineering analyses reveal that the embedding network primarily generates heterogeneous negative mean inputs that shift the core neurons into a higher-flux operating regime. Replacing the full time-dependent embedding signals by corresponding constant biases preserves this flux-enhancing effect, indicating that the detailed signal dynamics are not essential under the tested conditions. We further show that the information flux can be increased even beyond the biologically embedded case by applying optimized uniform or neuron-specific biases to the core neurons, and that similarly high values can emerge from a simple local adaptive-bias rule. Additional controls show that adaptive regulation and independently tuned inhibitory targeting further improve the embedded network, whereas removing the privileged strong-weight organization of the core abolishes the embedding advantage. Discussion: These results suggest that the surrounding network acts as a distributed regulator that prevents a strongly recurrent core from becoming trapped in saturated, low-flux states, rather than as an optimizer of information flux. Our findings are relevant both for the functional interpretation of biological neural circuits and for the design of artificial recurrent systems such as reservoir computers.
- 20 October 2026
- Published by Frontiers Media SA in Frontiers in Computational Neuroscience
- Vol. 20, 1886601
Open AccessAre cortical microcircuits optimized for information flux?—A simulation-based reverse engineering study
Introduction: A sufficiently large information flux in recurrent neural networks, quantified by the mutual information between successive network states, has been proposed as a marker of rich information-processing capabilities. This raises the question of whether biological neural networks, such as cortical microcircuits, may be structurally organized to enhance information flux. Methods: To investigate this possibility, we study a simplified model of the cortical layer 5 architecture, in which a densely and strongly interconnected core population is embedded within a larger supporting network. Results: Surprisingly, we find that the embedding network exerts a pronounced flux-enhancing effect on the core dynamics. Systematic reverse-engineering analyses reveal that the embedding network primarily generates heterogeneous negative mean inputs that shift the core neurons into a higher-flux operating regime. Replacing the full time-dependent embedding signals by corresponding constant biases preserves this flux-enhancing effect, indicating that the detailed signal dynamics are not essential under the tested conditions. We further show that the information flux can be increased even beyond the biologically embedded case by applying optimized uniform or neuron-specific biases to the core neurons, and that similarly high values can emerge from a simple local adaptive-bias rule. Additional controls show that adaptive regulation and independently tuned inhibitory targeting further improve the embedded network, whereas removing the privileged strong-weight organization of the core abolishes the embedding advantage. Discussion: These results suggest that the surrounding network acts as a distributed regulator that prevents a strongly recurrent core from becoming trapped in saturated, low-flux states, rather than as an optimizer of information flux. Our findings are relevant both for the functional interpretation of biological neural circuits and for the design of artificial recurrent systems such as reservoir computers.
- Open Access
- 15 October 2026
- Published by Frontiers Media SA in Frontiers in Complex Systems
- Vol. 4, 1920737
Surviving by Serving - the effects of rewarding usefulness
How can a complex production network organize itself when no participant knows the global plan? We investigate a simple principle inspired by markets: producers persist longer when what they make is actually used. We call this principle “Surviving by Serving” (SbS). In a deliberately minimal model, agents transform existing products into new ones and thereby build fluctuating production and supply networks starting from a single inexhaustible resource. We compare SbS with systems in which survival is not rewarded and with a control in which the same reward events are redirected to randomly selected agents. Within the tested model and parameter ranges, linking survival to downstream use enables faster and broader exploration of product space while retaining a larger share of what has been discovered. It also gives rise to more persistent, repeatedly active production pathways and more pronounced cyclic structures. In a targeted-reward experiment, selectively supporting suppliers of designated products can increase delivery to the desired region, without specifying the supply chains needed to reach it. These results do not establish a universal law of self-organization. They show, however, that a minimal local feedback between usefulness and persistence can substantially reshape collective organization in this model and motivate tests of analogous mechanisms in richer economic and biological networks.
- 15 October 2026
- Published by Frontiers Media SA in Frontiers in Complex Systems
- Vol. 4, 1920737
Open AccessSurviving by Serving - the effects of rewarding usefulness
How can a complex production network organize itself when no participant knows the global plan? We investigate a simple principle inspired by markets: producers persist longer when what they make is actually used. We call this principle “Surviving by Serving” (SbS). In a deliberately minimal model, agents transform existing products into new ones and thereby build fluctuating production and supply networks starting from a single inexhaustible resource. We compare SbS with systems in which survival is not rewarded and with a control in which the same reward events are redirected to randomly selected agents. Within the tested model and parameter ranges, linking survival to downstream use enables faster and broader exploration of product space while retaining a larger share of what has been discovered. It also gives rise to more persistent, repeatedly active production pathways and more pronounced cyclic structures. In a targeted-reward experiment, selectively supporting suppliers of designated products can increase delivery to the desired region, without specifying the supply chains needed to reach it. These results do not establish a universal law of self-organization. They show, however, that a minimal local feedback between usefulness and persistence can substantially reshape collective organization in this model and motivate tests of analogous mechanisms in richer economic and biological networks.
- Open Access
- 30 September 2026
- Published by Frontiers Media SA in Frontiers in Psychology
- Vol. 17, 1950309
Can artificial systems become conscious?—Recursive embodied closure as an integrative framework
Recent advances in artificial intelligence, connectomics, and computational neuroscience increasingly raise the question of whether artificial systems—or even pure computer simulations—could become conscious. Contemporary theories emphasize embodiment, predictive processing, homeostatic regulation, and continuous interaction between cognition, body, and environment rather than isolated computation. We propose recursive embodied closure as an integrative framework linking embodied, predictive, functionalist, and dynamical approaches to consciousness. Rather than replacing existing theories, the framework identifies a shared organizational principle: the continuous recursive coupling of cognition, bodily regulation, action, and environmental feedback. We argue that consciousness may, in principle, emerge in artificial systems, including computational simulations, for three reasons. First, consciousness may depend on recursively closed loops linking cognition, body, and environment, and such loops could also be realized in robotic, neuromorphic, or virtual embodiments. Second, evidence from comparative neuroscience, convergent evolution, neurotechnology, and philosophy of mind suggests that the mechanisms relevant for consciousness are largely substrate-independent and may therefore be implemented in both biological and artificial systems. Third, although unknown physical mechanisms cannot be excluded, there is currently no empirical evidence for biological processes that are uniquely necessary for consciousness. Taken together, these considerations support the view that consciousness is best understood as an emergent property of recursively embodied, environmentally embedded cognitive systems rather than of biological matter itself. If so, biological brains, artificial neural networks, neuromorphic hardware, and even virtual agents may constitute plausible candidates for future artificial consciousness.
- 30 September 2026
- Published by Frontiers Media SA in Frontiers in Psychology
- Vol. 17, 1950309
Open AccessCan artificial systems become conscious?—Recursive embodied closure as an integrative framework
Recent advances in artificial intelligence, connectomics, and computational neuroscience increasingly raise the question of whether artificial systems—or even pure computer simulations—could become conscious. Contemporary theories emphasize embodiment, predictive processing, homeostatic regulation, and continuous interaction between cognition, body, and environment rather than isolated computation. We propose recursive embodied closure as an integrative framework linking embodied, predictive, functionalist, and dynamical approaches to consciousness. Rather than replacing existing theories, the framework identifies a shared organizational principle: the continuous recursive coupling of cognition, bodily regulation, action, and environmental feedback. We argue that consciousness may, in principle, emerge in artificial systems, including computational simulations, for three reasons. First, consciousness may depend on recursively closed loops linking cognition, body, and environment, and such loops could also be realized in robotic, neuromorphic, or virtual embodiments. Second, evidence from comparative neuroscience, convergent evolution, neurotechnology, and philosophy of mind suggests that the mechanisms relevant for consciousness are largely substrate-independent and may therefore be implemented in both biological and artificial systems. Third, although unknown physical mechanisms cannot be excluded, there is currently no empirical evidence for biological processes that are uniquely necessary for consciousness. Taken together, these considerations support the view that consciousness is best understood as an emergent property of recursively embodied, environmentally embedded cognitive systems rather than of biological matter itself. If so, biological brains, artificial neural networks, neuromorphic hardware, and even virtual agents may constitute plausible candidates for future artificial consciousness.
- Open Access
- 14 August 2026
- Published by Frontiers Media SA in Frontiers in Computational Neuroscience
- Vol. 20, 1845838
Structural and dynamical strategies to prevent runaway excitation in reservoir computing
Reservoirs, typically implemented as recurrent neural networks (RNNs) with fixed random connection weights, can be combined with a simple trained readout layer to perform a wide range of computational tasks. However, increasing the magnitude of reservoir connection weights to exploit non-linear dynamics can cause the network to develop strong spontaneous activity that drives neurons into saturation, dramatically degrading performance. In this work, we investigate two distinct countermeasures against such runaway excitation. The first approach introduces a subtle non-homogeneous structure into the matrix of connection weights. w ij , without altering the overall probability distribution p ( w ). We identify several favorable structuring principles, such as creating a small subset of neurons with weaker-than-average input connections. Even if the rest of the reservoir falls into runaway saturating behavior, this weakly coupled subset remains in a mildly non-linear regime whose dynamics can still be exploited by the readout layer. The second approach implements a form of automatic gain control (AGC), in which a dedicated control unit dynamically regulates the reservoir's average global activation toward an optimal setpoint. Although the control unit modulates the excitability of the reservoir only via a global gain factor, this mechanism substantially enlarges the dynamical regime favorable for computation and renders performance largely independent of the underlying connection statistics.
- 14 August 2026
- Published by Frontiers Media SA in Frontiers in Computational Neuroscience
- Vol. 20, 1845838
Open AccessStructural and dynamical strategies to prevent runaway excitation in reservoir computing
Reservoirs, typically implemented as recurrent neural networks (RNNs) with fixed random connection weights, can be combined with a simple trained readout layer to perform a wide range of computational tasks. However, increasing the magnitude of reservoir connection weights to exploit non-linear dynamics can cause the network to develop strong spontaneous activity that drives neurons into saturation, dramatically degrading performance. In this work, we investigate two distinct countermeasures against such runaway excitation. The first approach introduces a subtle non-homogeneous structure into the matrix of connection weights. w ij , without altering the overall probability distribution p ( w ). We identify several favorable structuring principles, such as creating a small subset of neurons with weaker-than-average input connections. Even if the rest of the reservoir falls into runaway saturating behavior, this weakly coupled subset remains in a mildly non-linear regime whose dynamics can still be exploited by the readout layer. The second approach implements a form of automatic gain control (AGC), in which a dedicated control unit dynamically regulates the reservoir's average global activation toward an optimal setpoint. Although the control unit modulates the excitability of the reservoir only via a global gain factor, this mechanism substantially enlarges the dynamical regime favorable for computation and renders performance largely independent of the underlying connection statistics.
- Open Access
- 10 August 2026
- Published by Frontiers Media SA in Frontiers in Audiology and Otology
- Vol. 4, 1910758
Ten years of the stochastic resonance model of tinnitus: from phantom perception to adaptive sensory optimization
Subjective tinnitus—the perception of sound in the absence of an external acoustic stimulus—remains one of the most debated phenomena in auditory neuroscience. In 2016, the stochastic resonance (SR) model was introduced as an alternative account of tinnitus-related neuronal hyperactivity, proposing that internally generated neural noise is adaptively upregulated to restore information transmission after hearing loss. Rather than interpreting increased spontaneous activity as purely maladaptive, the model reframed it as a functional mechanism that enhances signal detection near sensory thresholds, with tinnitus emerging as a perceptual side effect of adaptive sensory optimization. Over the past decade, this framework has evolved from a phenomenological hypothesis into a broader neuro-computational theory integrating information theory, adaptive signal detection, multichannel auditory processing, and cross-modal plasticity. The available literature spans different levels of evidence: computational simulations establish proof of principle, whereas animal experiments, observational analyses, and early clinical studies test distinct and partially overlapping predictions of the model. The framework has also motivated a specific parameterization of sound-based interventions using spectrally matched near-threshold noise. Established tinnitus maskers and sound therapies were developed independently of the SR model. Their comparative clinical efficacy has yet to be systematically evaluated. Recently, the SR framework has been embedded within a unified theory of auditory (phantom) perception that integrates stochastic resonance, central gain, homeostatic plasticity, and predictive coding. This review provides a chronological account of the development of the stochastic resonance model of tinnitus, summarizes theoretical refinements and empirical evidence, and outlines future directions for mechanistic validation and clinical translation. By redefining tinnitus as an emergent consequence of adaptive sensory computation, the model shifts the conceptual focus from pathological dysfunction to principles of information optimization in neural systems.
- 10 August 2026
- Published by Frontiers Media SA in Frontiers in Audiology and Otology
- Vol. 4, 1910758
Open AccessTen years of the stochastic resonance model of tinnitus: from phantom perception to adaptive sensory optimization
Subjective tinnitus—the perception of sound in the absence of an external acoustic stimulus—remains one of the most debated phenomena in auditory neuroscience. In 2016, the stochastic resonance (SR) model was introduced as an alternative account of tinnitus-related neuronal hyperactivity, proposing that internally generated neural noise is adaptively upregulated to restore information transmission after hearing loss. Rather than interpreting increased spontaneous activity as purely maladaptive, the model reframed it as a functional mechanism that enhances signal detection near sensory thresholds, with tinnitus emerging as a perceptual side effect of adaptive sensory optimization. Over the past decade, this framework has evolved from a phenomenological hypothesis into a broader neuro-computational theory integrating information theory, adaptive signal detection, multichannel auditory processing, and cross-modal plasticity. The available literature spans different levels of evidence: computational simulations establish proof of principle, whereas animal experiments, observational analyses, and early clinical studies test distinct and partially overlapping predictions of the model. The framework has also motivated a specific parameterization of sound-based interventions using spectrally matched near-threshold noise. Established tinnitus maskers and sound therapies were developed independently of the SR model. Their comparative clinical efficacy has yet to be systematically evaluated. Recently, the SR framework has been embedded within a unified theory of auditory (phantom) perception that integrates stochastic resonance, central gain, homeostatic plasticity, and predictive coding. This review provides a chronological account of the development of the stochastic resonance model of tinnitus, summarizes theoretical refinements and empirical evidence, and outlines future directions for mechanistic validation and clinical translation. By redefining tinnitus as an emergent consequence of adaptive sensory computation, the model shifts the conceptual focus from pathological dysfunction to principles of information optimization in neural systems.
- Open Access
- 6 August 2026
- Published by Frontiers Media SA in Frontiers in Language Sciences
- Vol. 5, 1846406
Convergent representations of linguistic constructions in human and artificial neural systems
Understanding how the brain processes linguistic constructions is a central challenge in cognitive neuroscience and linguistics. Recent computational studies have shown that artificial neural language models spontaneously develop internally differentiated representations of Argument Structure Constructions (ASCs), generating specific predictions about when and how construction-level information should emerge during language processing. The present study tests these predictions in human neural activity using electroencephalography (EEG). Data were recorded from 12 native English speakers, with 11 participants retained for all reported analyses after preprocessing. Construction-related activity was analyzed using time-frequency analysis, statistical feature extraction, and machine learning classification. The results reveal construction-specific neural signatures that emerge primarily at sentence-final positions, where argument structure information becomes fully disambiguated, and are most prominent in the alpha frequency band. Pairwise classification further demonstrated reliable differentiation between specific construction pairs, particularly ditransitive and resultative constructions, while other pairs exhibited overlapping neural signatures. Importantly, the temporal emergence and similarity structure of these neural effects closely mirror patterns previously observed in recurrent and transformer-based language models, in which constructional representations arise during integrative stages of processing. Together, these findings provide converging evidence that linguistic constructions are neurally encoded as distinct form-meaning mappings, supporting core assumptions of Construction Grammar and suggesting that both biological and artificial learning systems converge on similar representational solutions during predictive language processing. More broadly, this convergence is consistent with the idea that learning systems may discover stable regions within an underlying representational landscape - recently referred to as a Platonic representational space - that constrains which linguistic abstractions emerge as cognitively and computationally efficient forms.
- 6 August 2026
- Published by Frontiers Media SA in Frontiers in Language Sciences
- Vol. 5, 1846406
Open AccessConvergent representations of linguistic constructions in human and artificial neural systems
Understanding how the brain processes linguistic constructions is a central challenge in cognitive neuroscience and linguistics. Recent computational studies have shown that artificial neural language models spontaneously develop internally differentiated representations of Argument Structure Constructions (ASCs), generating specific predictions about when and how construction-level information should emerge during language processing. The present study tests these predictions in human neural activity using electroencephalography (EEG). Data were recorded from 12 native English speakers, with 11 participants retained for all reported analyses after preprocessing. Construction-related activity was analyzed using time-frequency analysis, statistical feature extraction, and machine learning classification. The results reveal construction-specific neural signatures that emerge primarily at sentence-final positions, where argument structure information becomes fully disambiguated, and are most prominent in the alpha frequency band. Pairwise classification further demonstrated reliable differentiation between specific construction pairs, particularly ditransitive and resultative constructions, while other pairs exhibited overlapping neural signatures. Importantly, the temporal emergence and similarity structure of these neural effects closely mirror patterns previously observed in recurrent and transformer-based language models, in which constructional representations arise during integrative stages of processing. Together, these findings provide converging evidence that linguistic constructions are neurally encoded as distinct form-meaning mappings, supporting core assumptions of Construction Grammar and suggesting that both biological and artificial learning systems converge on similar representational solutions during predictive language processing. More broadly, this convergence is consistent with the idea that learning systems may discover stable regions within an underlying representational landscape - recently referred to as a Platonic representational space - that constrains which linguistic abstractions emerge as cognitively and computationally efficient forms.
- Open Access
- 19 May 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1)
Illuminating the black box of reservoir computing
Reservoir computers, using recurrent neural networks with fixed random connections, are known to perform a wide range of information-processing tasks. Yet the transformations taking place within the reservoir, the interaction between input matrix, reservoir, and readout layer, and the influence of key design parameters remain insufficiently understood. Here, we shift the focus from performance maximization to the identification of minimal computational requirements for different model tasks. We investigate how many neurons and how much nonlinearity are needed to solve specific tasks, including cases with non-sigmoidal activation functions. Our results show that the division of labor between input matrix, reservoir, and readout layer depends strongly on the task. In many cases, a weakly coupled and only minimally nonlinear reservoir proves sufficient. In addition, features often considered secondary, such as the structure of the input matrix or the steepness of the activation functions, can become decisive for performance.
- 19 May 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1)
Open AccessIlluminating the black box of reservoir computing
Reservoir computers, using recurrent neural networks with fixed random connections, are known to perform a wide range of information-processing tasks. Yet the transformations taking place within the reservoir, the interaction between input matrix, reservoir, and readout layer, and the influence of key design parameters remain insufficiently understood. Here, we shift the focus from performance maximization to the identification of minimal computational requirements for different model tasks. We investigate how many neurons and how much nonlinearity are needed to solve specific tasks, including cases with non-sigmoidal activation functions. Our results show that the division of labor between input matrix, reservoir, and readout layer depends strongly on the task. In many cases, a weakly coupled and only minimally nonlinear reservoir proves sufficient. In addition, features often considered secondary, such as the structure of the input matrix or the steepness of the activation functions, can become decisive for performance.
- Open Access
- 30 April 2026
- Published by Elsevier in NeuroImage
- Vol. 334, 121966
The predictive brain: Neural correlates of word expectancy align with large language model prediction probabilities
- 30 April 2026
- Published by Elsevier in NeuroImage
- Vol. 334, 121966
Open AccessThe predictive brain: Neural correlates of word expectancy align with large language model prediction probabilities
- Open Access
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-22
Mind in the Machine? Cross-Disciplinary Perceptions of Consciousness in Artificial Intelligence
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-22
Open AccessMind in the Machine? Cross-Disciplinary Perceptions of Consciousness in Artificial Intelligence