Tiny dLIF: A Dendritic Spiking Neural Network Enabling a Time-Domain Energy-Efficient Seizure Detection System DOI Creative Commons
Luis Fernando Herbozo Contreras, Leping Yu, Zhaojing Huang

et al.

medRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: May 24, 2024

Abstract Epilepsy poses a significant global health challenge, driving the need for reliable diagnostic tools like scalp electroencephalogram (EEG), subscalp EEG, and intracranial EEG (iEEG) accurate seizure detection, localization, modulation treating seizures. However, these techniques often rely on feature extraction such as Short Time Fourier Transform (STFT) efficiency in detection. Drawing inspiration from brain architecture, we investigate biologically plausible algorithms, specifically emphasizing time-domain inputs with low computational overhead. Our novel approach features two hidden layer dendrites Leaky Integrate-and-Fire (dLIF) spiking neurons, containing fewer than 300K parameters occupying mere 1.5 MB of memory. proposed network is tested successfully generalized four datasets USA Europe, recorded different front-end electronics. are adults children, European iEEG adults. All patients living epilepsy. model exhibits robust performance across through rigorous training validation. We achieved AUROC scores 81.0% 91.0% datasets. Additionally, obtained AUPRC F1 Score metrics 91.9% 88.9% one dataset, respectively. also conducted out-of-sample generalization by adult patient data, testing children’s achieving an 75.1% epilepsy This highlights its effectiveness continental diverse modalities, regardless montage or age specificity. It underscores importance embracing system heterogeneity to enhance efficiency, thus eliminating computationally expensive engineering Fast (FFT) STFT.

Language: Английский

Bio-plausible reconfigurable spiking neuron for neuromorphic computing DOI Creative Commons
Xiao Yu, Y. F. Liu,

Bihua Zhang

et al.

Science Advances, Journal Year: 2025, Volume and Issue: 11(6)

Published: Feb. 5, 2025

Biological neurons use diverse temporal expressions of spikes to achieve efficient communication and modulation neural activities. Nonetheless, existing neuromorphic computing systems mainly simplified neuron models with limited spiking behaviors due high cost emulating these biological spike patterns. Here, we propose a compact reconfigurable design using the intrinsic dynamics NbO 2 -based unit excellent tunability in an electrochemical memory (ECRAM) emulate fast-slow bio-plausible neuron. The resistance ECRAM was effective tuning membrane potential, contributing flexible reconfiguration various firing modes, such as phasic burst spiking, exhibiting adaptive changing environment. We used model build networks bursting demonstrated improved classification accuracies over models, showing great promises for more systems.

Language: Английский

Citations

2
Leveraging dendritic properties to advance machine learning and neuro-inspired computing DOI Creative Commons
Michalis Pagkalos,

Roman Makarov,

Panayiota Poirazi

et al.

Current Opinion in Neurobiology, Journal Year: 2024, Volume and Issue: 85, P. 102853 - 102853

Published: Feb. 22, 2024

The brain is a remarkably capable and efficient system. It can process store huge amounts of noisy unstructured information using minimal energy. In contrast, current artificial intelligence (AI) systems require vast resources for training while still struggling to compete in tasks that are trivial biological agents. Thus, brain-inspired engineering has emerged as promising new avenue designing sustainable, next-generation AI systems. Here, we describe how dendritic mechanisms neurons have inspired innovative solutions significant problems, including credit assignment multilayer networks, catastrophic forgetting, high energy consumption. These findings provide exciting alternatives existing architectures, showing research pave the way building more powerful energy-efficient learning

Language: Английский

Citations

13
Winter Wheat Yield Prediction Using Satellite Remote Sensing Data and Deep Learning Models DOI Creative Commons

Hongkun Fu,

Jian Lü,

Jian Li

et al.

Agronomy, Journal Year: 2025, Volume and Issue: 15(1), P. 205 - 205

Published: Jan. 16, 2025

Accurate crop yield prediction is crucial for formulating agricultural policies, guiding management, and optimizing resource allocation. This study proposes a method predicting yields in China’s major winter wheat-producing regions using MOD13A1 data deep learning model which incorporates an Improved Gray Wolf Optimization (IGWO) algorithm. By adjusting the key parameters of Convolutional Neural Network (CNN) with IGWO, accuracy significantly enhanced. Additionally, explores potential Green Normalized Difference Vegetation Index (GNDVI) prediction. The research utilizes collected from March to May between 2001 2010, encompassing vegetation indices, environmental variables, statistics. results indicate that IGWO-CNN outperforms traditional machine approaches standalone CNN models terms accuracy, achieving highest performance R2 0.7587, RMSE 593.6 kg/ha, MAE 486.5577 MAPE 11.39%. finds April optimal period early wheat. validates effectiveness combining remote sensing prediction, providing technical support precision agriculture contributing global food security sustainable development.

Language: Английский

Citations

1
Bionic Recognition Technologies Inspired by Biological Mechanosensory Systems DOI Open Access
Xiangxiang Zhang, Chang-Guang Wang, Xin Pi

et al.

Advanced Materials, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 21, 2025

Abstract Mechanical information is a medium for perceptual interaction and health monitoring of organisms or intelligent mechanical equipment, including force, vibration, sound, flow. Researchers are increasingly deploying recognition technologies (MIRT) that integrate acquisition, pre‐processing, processing functions expected to enable advanced applications. However, this also poses significant challenges acquisition performance efficiency. The novel exciting mechanosensory systems in nature have inspired us develop superior bionic (MIBRT) based on materials, structures, devices address these challenges. Herein, first strategies pre‐processing presented their importance high‐performance highlighted. Subsequently, design considerations sensors by mechanoreceptors described. Then, the concepts neuromorphic summarized order replicate biological nervous system. Additionally, ability MIBRT investigated recognize basic information. Furthermore, further potential applications robots, healthcare, virtual reality explored with view solve range complex tasks. Finally, future opportunities identified from multiple perspectives.

Language: Английский

Citations

1
Uncertainty-Aware Graph Contrastive Fusion Network for multimodal physiological signal emotion recognition DOI
Guangqiang Li, Ning Chen, Hongqing Zhu

et al.

Neural Networks, Journal Year: 2025, Volume and Issue: 187, P. 107363 - 107363

Published: March 14, 2025

Language: Английский

Citations

1
Cardiac Heterogeneity Prediction by Cardio-Neural Network Simulation DOI
Asif Mehmood,

Ayesha Ilyas,

H. Ilyas

et al.

Neuroinformatics, Journal Year: 2025, Volume and Issue: 23(2)

Published: Feb. 1, 2025

Language: Английский

Citations

0
Msfi: Multi-Timescale Spatio-Temporal Features Integration in Spiking Neural Networks DOI
Dengfeng Xue, Wenjuan Li, Chunfeng Yuan

et al.

Published: Jan. 1, 2025

Language: Английский

Citations

0
Complex Spiking Neural Network Evaluated by Injury Resistance Under Stochastic Attacks DOI Creative Commons
Lei Guo, Chang Ming Li, Huan Liu

et al.

Brain Sciences, Journal Year: 2025, Volume and Issue: 15(2), P. 186 - 186

Published: Feb. 13, 2025

Brain-inspired models are commonly employed for artificial intelligence. However, the complex environment can hinder performance of electronic equipment. Therefore, enhancing injury resistance brain-inspired is a crucial issue. Human brains have self-adaptive abilities under injury, so drawing on advantages human brain to construct model intended enhance its resistance. But current still lack bio-plausibility, meaning they do not sufficiently draw real neural systems' structure or function. To address this challenge, paper proposes spiking network (Com-SNN) as model, in which topology inspired by topological characteristics biological functional networks, nodes Izhikevich neuron models, and edges synaptic plasticity with time delay co-regulated excitatory synapses inhibitory synapses. evaluate Com-SNN, two injury-resistance metrics investigated compared SNNs alternative topologies stochastic removal simulate consequence attacks. In addition, mechanism remains unclear, revealing understanding development analyzes dynamic regulation Com-SNN The experimental results indicate that superior other SNNs, demonstrating our help improve SNNs. Our imply an intrinsic element impacting resistance, another impacts

Language: Английский

Citations

0
Flexible Co-TCPP Nanosheet-Based Memristor for Neuromorphic Computing and Simulation of Human Water Turnover at Different Temperatures DOI

Guoyao Ouyang,

Yilong Wang, Jie Su

et al.

Nano Energy, Journal Year: 2025, Volume and Issue: unknown, P. 110778 - 110778

Published: Feb. 1, 2025

Language: Английский

Citations

0
The effectiveness evaluation of industry education integration model for applied universities under back propagation neural network DOI Creative Commons

Ying Qi,

Wei Feng

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: Feb. 15, 2025

As the education field continues to advance, industry–education integration has become a crucial strategy for enhancing teaching quality in applied universities. This study investigates how artificial intelligence, specifically back propagation neural network (BPNN), can be within an framework strengthen students' skills and employability. A series of experiments were conducted assess model's effectiveness linking theoretical learning with practical experience, as well improving hands-on innovative abilities. Results demonstrate that BPNN-optimized model substantially boosts overall competencies. For instance, average academic score students experimental group rose from 78.5 85.2, assessment scores increased 76.8 88.4, innovation improved 74.2 82.5. Additionally, employment rate reached 94%, surpassing control group's 76%, significant gains job satisfaction career planning skills. These findings highlight BPNN-based effectively strengthens knowledge, skills, employability, offering valuable enhanced university-industry collaboration.

Language: Английский

Citations

0