The Gain-Control property of Synaptic Dynamics (SD) describes how the response of a neuron, mediated by short-term depression, can be sensitive to proportional changes of stochastic inputs.This has been related to the temporal filtering property of synaptic efficacy.However, the limits of the strength of Gain-Control in relationship to the efficacy have not been explored.This paper meets this gap by simulating networks using two synapses with fast-and slow-decays of efficacy.The results show that the Gain-Control effect decreases with the decay of efficacy.Formally describing this relationship can facilitate the integration of SD into Spiking Neural Networks.

Despite the remarkable progress in the field of artificial intelligence (AI), current models seldom incorporate emotions or affective states, constraining their capacity for emulating truly adaptive and human-like behavior. Boredom is a recurrent affective state that signals a mismatch between available cognitive resources and environmental demands, prompting a state-escape response to restore optimal engagement. Cast within a functional psychological framework, boredom is characterized to arise when the level of cognitive engagement falls outside an optimal range, whether due to under- or overstimulation. Here, we translate this principle into a biologically inspired control loop implemented with spiking neural networks. The model continuously monitors simulated cognitive resource utilization, signals deviations that occur due to perturbations in the input and dynamically influences the utilization of resources to maintain an optimal engagement level. Simulations demonstrate that the model effectively maintains stable cognitive engagement by exciting and inhibiting a spiking neuron population that abstractly represents the processing of input. This work establishes a foundation towards the development of a future model capable of autonomously defining and regulating its optimal level of cognitive engagement. By embedding such affective regulation directly into a spiking architecture, our approach bridges cognitive neuroscience and AI, offering insights into how human-like state monitoring and initiating of corrective actions can be realized within brain-inspired AI.

Integrating the modulatory properties of Synaptic Dynamics (SD) into Spiking Neural Networks (SNNs) can enhance their computational capabilities.For improving this integration process, this paper presents a pipeline based on Differential Evolution to tune parameters of SD models.Using reference signals from in vitro experiments, parameters of two models are tuned as study cases: the Tsodyks-Markram and the Modified Stochastic Synaptic Model.The pipeline has an average success rate of 75% and 80% respectively.The outcome is a distribution of parameters for each model, which can be considered as prior knowledge to facilitate the integration of SD models into SNNs.

This work presents a novel Spiking Neural Network training architecture based on a deepened Liquid State Machine integrating Winner-Takes-All computation and Reward-Modulated Synaptic Plasticity.The networks performance is evaluated on the Heidelberg Dataset for spoken digit recognition.A two-layer liquid configuration improves classification accuracy by 5% over a single-layer baseline, while incorporating feedback between liquid layers.This architecture demonstrates that deep liquid models, combined with feedback and reward-driven learning, can effectively capture complex spatio-temporal patterns, offering significant advantages in terms of accuracy over traditional Liquid State Machines.

Boredom is increasingly recognized as a functional emotion playing an important role in regulating human behavior.Despite continuous advances in the field of artificial intelligence, research on whether these models can enter emotional states such as boredom remains limited.However, emotions can be pivotal towards more human-like intelligence in AI.This paper transfers the regulatory function of boredom into a control loop modeled with spiking neural networks.Simulations demonstrate the successful replication of the regulatory mechanism of boredom based on simulated input.This work provides a foundation for future research and development towards a self-regulating system based on spiking neural networks capable of entering a state of boredom.

The paper deals with the analysis of the sleep patterns of a patient with Completely Locked-In Syndrome (CLIS).The analysis was performed using an approach initially designed to detect consciousness in Disorders of Consciousness (DoC) and CLIS patients.The method extracts different features based on spectral, complexity and connectivity measures and performs soft-clustering analyses to determine the consciousness state.The results showed that it was able to discriminate between the (Non)-Rapid Eye Movement (NREM) and the Rapid Eye Movement (REM) sleep stages.Detecting normal Slow-Wave Sleep (SWS) and REM phases indicates better communication abilities for the patient.

In the past, the question regarding the point of singularity in artificial intelligence -when machines become more intelligent than humans -has been raised again and again.In this publication, a crucial point of human intelligence and the impact on this discussion will be postulated in the form of 3 hypotheses as thought-provoking impulse based on the basic hypothesis, that only systems which can be bored are intelligent.First, boredom is discussed from the perspective of psychology with its influence on human intelligence before deductions are drawn from this to artificial intelligence resp.machine learning.Finally, the hypotheses are formulated and the resulting future investigations are outlined.

Convolutional Neural Networks have become an important tool for various Computer Vision tasks. Yet, increasing complexity of such architectures drives computational costs. To this end, we propose two measures to achieve similar classification results as state-of-the-art architectures while at the same time reducing model complexity significantly. Firstly, we describe a novel type of non-linear parameter-efficient morphological layers inspired by concepts that are well-known and widely used with convolutions. Secondly, we present a set of simple network architectures, organized as optimization framework, which is enhanced by neural architecture search and hyperparameter optimization. In experiments with hyperspectral remote sensing data, we demonstrate that the identified optimal morphological architecture produces results not only comparable with other architectures from the optimization framework, but also comparable or better than selected state-of-the-art neural network architectures for image classification. Depending on the performed task, the proposed optimized architecture requires up to 25 times fewer parameters than actual state-of-the-art networks.

Field-programmable gate arrays (FPGAs) have re-cently received renewed attention in the context of Evolvable Hardware (EHW). The most fine grained approach to changing their internal structure, direct manipulation of the bitstream, has largely been abandoned. The undocumented bitstream formats of modern FPGAs made it complicated and error-prone. This situation has fundamentally changed with the advent of open-source FPGA toolchains. Previous attempts to exploit this opportunity were promising, but only manged to solve very basic tasks. We present in this paper an evolved tone discriminator circuit. It was evolved by replicating the most famous experiment in this field, but with modern hardware. For that we map the originally used Xilinx XC6200 FPGA to a modern Lattice iCE40 FPGA. We show how to set up the experiment and optimize the evolution environment. Our approach allows over 130 times more reconfigurations per second than previous approaches. Additionally, we discuss reasons for the abandonment of direct bitstream manipulation for EHW in context of the new possibilities created by open-source FPGA toolchains. We show which challenges have been solved and which steps need to be taken next.

Reducing the bitstream size is important to lower external storage requirements and to speed-up the reconfiguration of field-programmable gate arrays (FPGAs). The most common methods for bitstream size reduction are based on dedicated hardware elements or dynamic partial reconfiguration. All of these properties are usually missing in low-cost FPGAs such as the Lattice iCE40 device family. In this paper we propose a lightweight compaction approach for iCE40 FPGAs. We present five methods for bitstream compaction: two adapted and three new. The methods work directly on the bitstream by removing unnecessary data and redundant commands. They are applicable independent of the synthesis toolchain and require neither repetition of synthesis steps nor modifications of the target system. Although our focus is on iCE40 devices, we additionally discuss the conditions for applying our approach to other targets. All five methods were implemented in an open-source compaction tool. We evaluate our approach with an iCE40 HX8K FPGA by synthesizing and compacting various projects. As a result, we achieve a reduction in bitstream size and reconfiguration time by up to 79 %.

Objectives: The aim of this investigation is to study development of the functional connectivity and the complexity of the electroencephalogram (EEG) recorded from completely locked-in patients (CLIS) across two consecutive years. CLIS patients lose their ability to voluntary control their muscles but maintain cognitive abilities, although the extant is unknown. The goal is to investigate the effect of time on brain signals connectivity and complexity in the completely locked-in state over time. Methods: EEG data were recorded from three CLIS patients, identified by the numbers 2, 5 and 7. Each of the patients satisfy the following conditions: at least two years of available recordings, specifically signals from channels FC5 and FC6. This results in two years data for each of the patients. Since the recording channels were not consistent over the years, the choice of the channels was made on the basis that they were common to the most of these patients for more than a year, for comparability purposes. The connectivity measures were evaluated using the imaginary part of the coherence (iCOH) and the weighted symbolic mutual information (wSMI) on the recorded data. Using only the imaginary part of the coherence allows to reduce the effects of volume conduction. On the other hand, wSMI values estimate the information shared between two EEG channels. In both method, a higher value suggests an increase of the functional connectivity between two channels or two brain regions. The complexity was estimated using the Poincaré index, computed as the ratio between the short-term variability and the long-term variability of the Poincaré plot. A high Poincaré index value suggests an activated cortex. Results: Imaginary coherence and wSMI values showed high variability from day to day the first year for Patients 2 and 5. The overall connectivity values slightly decrease from the first to the second year for both patients. Similarly, for both patients, slight increases of the Poincaré index values are observed from the first to the second year. As for Patient 7, coherence values were in the same range for both years as opposed to the wSMI and Poincaré index values, which showed substantial differences. Conclusion: Patients 2 and 5 showed similar results, indicating a possible degradation of the functional connectivity between the chosen channels but an increased activity at the single channel level, on the contrary of Patient 7. The decrease in the connectivity values suggests a progressive decrease of communication between the channels located in the motor area, correlated with the loss of residual invisible motor function.

An important aspect of clinical work, especially in the intensive care unit (ICU) is the vigilant observation of patients' vital data on a various number of monitors and displays to allow for rapid decision-making which is often required in critical conditions. Most often, the presented data cannot be collected and used as a whole to aid the decision making by availability and comparison of all relevant data at once. In this paper, a novel approach for handling real-time clinical data is presented. We developed TICoMS (Tübinger ICU Control and Monitoring System), a framework for integration of sensors and effectors into a homogeneous message-based network capable of gathering and distributing real-time data in ICU settings from medical devices. The accumulated data can be presented on a central patient monitoring screen for a fast and extensive overview of the patient's current condition and trends with a homogenous time-base. Furthermore, this framework provides control capabilities for medical devices and is therefore an essential foundation for device interactions that may allow for computer-based support in decision-making or advanced automation in intensive care scenarios.

We outline a blueprint for the development of new computer science approaches for the management and analysis of big data problems for biodiversity science. Such problems are characterized by a combination of different data sources each of which owns at least one of the typical characteristics of big data (volume, variety, velocity, or veracity). For these problems, we envision a solution that covers different aspects of integrating data sources and algorithms for their analysis on one of the following three layers: At the data layer, there are various data archives of heterogeneous, unstructured, and uncertain data. At the functional layer, the data are analyzed for each archive individually. At the meta-layer, multiple functional archives are combined for complex analysis.

According to Cognitive Load Theory the type and amount of workload (WL) during learning is crucial for successful learning and should be held within an optimal range of learners' memory capacity. Therefore, we aim at developing electroencephalogram (EEG) based learning environments adapting to learners individual WL online. To achieve this goal efficient classification methods are necessary. Support Vector Machines (SVMs) can accurately classify WL using within-task classification, but within-task classification is not feasible in complex learning environments. Therefore, the present study examined cross-task classification accuracies for SVMs trained on EEG-signals, recorded while participants (N= 21) had to solve three working memory tasks. While within-task classification accuracies were high for WM tasks (average: 95% - 97 %), cross-task classification performances were not significant over chance level. Since cross-task classification is a necessary step towards developing generalized classifiers, we will discuss the benefits and drawbacks as well as possible enhancements in the course of this paper to use it as an effective approach for learning environments.

Electrocorticography (ECoG) offers the possibility of decoding movement intention even in the absence of motor control, making it a powerful signal source for brain-computer interfaces (BCI). We designed a BCI that translates attempts to move the hand into movements of a video of an opening hand to investigate its use for pain therapy and stroke rehabilitation. One patient with phantom limb pain after amputation of the arm and one patient suffering from chronic pain and paralysis after a stroke trained with this BCI for several sessions. Signals were acquired with epidural ECoG grids placed over the motor cortex contralateral to the affected or missing hand. The analysis of data obtained in screening sessions with cued attempted movements showed highly significant (p2values for the discrimination between movement and rest conditions for most frequencies up to 200 Hz. Both patients acquired control of the BCI system which was verified by the evaluation of three measures of the ability to start and stop the video application. In particular, both patients learned to reliably start the video application in all trials. This demonstrates that it is feasible for patients with phantom limb pain and chronic pain as well as paralysis after stroke to operate a BCI that targets their missing or impaired limb, making it a potentially useful tool for new approaches in pain therapy and stroke rehabilitation.

The task of automatically tracking the visual attention in dynamic visual scenes is highly challenging. To approach it, we propose a Bayesian online learning algorithm. As the visual scene changes and new objects appear, based on a mixture model, the algorithm can identify and tell visual saccades (transitions) from visual fixation clusters (regions of interest). The approach is evaluated on real-world data, collected from eye-tracking experiments in driving sessions.

The visual input is perhaps the most important sensory information. Understanding its mechanisms as well as the way visual attention arises could be highly beneficial for many tasks involving the analysis of users' interaction with their environment. We present Vishnoo (Visual Search Examination Tool), an integrated framework that combines configurable search tasks with gaze tracking capabilities, thus enabling the analysis of both, the visual field and the visual attention. Our user studies underpin the viability of such a platform. Vishnoo is an open-source software and is available for download at http://www.vishnoo.de/

Kohonen's self-organizing map (SOM) is a popular neural network architecture for solving problems in the field of explorative data analysis, clustering, and data visualization. One of the major drawbacks of the SOM algorithm is the difficulty for nonexpert users to interpret the information contained in a trained SOM. In this paper, this problem is addressed by introducing an enhanced version of the Clusot algorithm. This algorithm consists of two main steps: 1) the computation of the Clusot surface utilizing the information contained in a trained SOM and 2) the automatic detection of clusters in this surface. In the Clusot surface, clusters present in the underlying SOM are indicated by the local maxima of the surface. For SOMs with 2-D topology, the Clusot surface can, therefore, be considered as a convenient visualization technique. Yet, the presented approach is not restricted to a certain type of 2-D SOM topology and it is also applicable for SOMs having an n-dimensional grid topology.

The aim of this paper is to show that machine learning techniques can be used to derive a classifying function for human brain signal data measured by magnetoencephalography (MEG), for the use in a brain computer interface (BCI). This is especially helpful for evaluating quickly whether a BCI approach based on electroencephalography, on which training may be slower due to lower signal-to-noise ratio, is likely to succeed. We apply RCE and regularized SVMs to the experimental data of ten healthy subjects performing a motor imagery task. Four subjects were able to use a trained classifier to write a short name. Further analysis gives evidence that the proposed imagination task is suboptimal for the possible extension to a multiclass interface. To the best of our knowledge this paper is the first working online MEG-based BCI and is therefore a "proof of concept".

This work presents the whole system-on-silicon design flow using systemC system specification language. In this study, systemC is used to design a multilayer perceptron neural network, which is applied to an electrocardiogram pattern recognition system. The objective of this work is to exemplify the synthesis of RTL-and behavioral integrated systems. To achieve this, a preprocessing methodology was used to optimize the three main constraints of hardware neural network (HNN) design: accuracy, space and processing speed. This allows a complex HNN to be implemented on a single field programmable gate array (FPGA). The high level systemC synthesis allows the straightforward translation of system level into hardware level, avoiding the error prone and the time consuming translation into another hardware description language.

Two artificial neural net (ANN) based controllers using interfaces to the peripheral nervous system are presented. The aim of the paper is to show the ANNs' capability to interact with the biological nervous system for control purposes. First, we present a system for controlling a limb prostheses by means of biological nerve signals. Recordings of nerve signals done by regeneration type neurosensors interfacing the peripheral nerve system are processed by ANNs in order to control a commercial available hand prostheses. The second system presented is the inverse of the presented controller above. In this case, the aim of the controller is the restoration of lost hand function for disabled persons Here, the ANN is the kernel of a closed loop control calculating stimulation patterns for the peripheral nerve system in order to evoke hand movements related to the patient's intent.