Training Spiking Neural Networks Using Lessons From Deep Learning

Proceedings of the IEEE, Год журнала: 2023, Номер 111(9), С. 1016 - 1054

Опубликована: Сен. 1, 2023

The brain is the perfect place to look for inspiration develop more efficient neural networks. inner workings of our synapses and neurons provide a glimpse at what future deep learning might like. This article serves as tutorial perspective showing how apply lessons learned from several decades research in learning, gradient descent, backpropagation, neuroscience biologically plausible spiking networks (SNNs). We also explore delicate interplay between encoding data spikes process; challenges solutions applying gradient-based SNNs; subtle link temporal backpropagation spike timing-dependent plasticity; move toward online learning. Some ideas are well accepted commonly used among neuromorphic engineering community, while others presented or justified first time here. A series companion interactive tutorials complementary this using Python package, snnTorch , made available: https://snntorch.readthedocs.io/en/latest/tutorials/index.html.

Язык: Английский

---

Intelligent Computing Technique to Analyze the Two-Phase Flow of Dusty Trihybrid Nanofluid with Cattaneo-Christov Heat Flux Model Using Levenberg-Marquardt Neural-Networks

Case Studies in Thermal Engineering, Год журнала: 2025, Номер unknown, С. 105891 - 105891

Опубликована: Фев. 1, 2025

Язык: Английский

---

Computational AI to predict and optimize the relationship between dye removal efficiency and Gibbs free energy in the adsorption process utilizing TiO2/chitosan-polyacrylamide composite

International Journal of Biological Macromolecules, Год журнала: 2024, Номер 264, С. 130738 - 130738

Опубликована: Март 7, 2024

Язык: Английский

---

Parametric Estimation of Directional Wave Spectra from Moored FPSO Motion Data Using Optimized Artificial Neural Networks

Journal of Marine Science and Engineering, Год журнала: 2025, Номер 13(1), С. 69 - 69

Опубликована: Янв. 3, 2025

This paper introduces a comprehensive, data-driven framework for parametrically estimating directional ocean wave spectra from numerically simulated FPSO (Floating Production Storage and Offloading) vessel motions. Leveraging mid-fidelity digital twin of spread-moored in the Guyana Sea, this approach integrates wide range statistical values calculated time histories responses—displacements, angular velocities, translational accelerations. Artificial neural networks (ANNs), trained optimized through hyperparameter tuning feature selection, are employed to estimate parameters including significant height, peak period, main direction, enhancement parameter, directional-spreading factor. A systematic correlation analysis ensures that informative input features retained, while extensive sensitivity tests confirm richer sets notably improve predictive accuracy. In addition, comparisons against other machine learning (ML) methods—such as Support Vector Machines, Random Forest, Gradient Boosting, Ridge Regression—demonstrate present ANN model’s superior ability capture intricate nonlinear interdependencies between motions environmental conditions.

Язык: Английский

---

Multiscale brain modeling: bridging microscopic and macroscopic brain dynamics for clinical and technological applications

Frontiers in Cellular Neuroscience, Год журнала: 2025, Номер 19

Опубликована: Фев. 19, 2025

The brain's complex organization spans from molecular-level processes within neurons to large-scale networks, making it essential understand this multiscale structure uncover brain functions and address neurological disorders. Multiscale modeling has emerged as a transformative approach, integrating computational models, advanced imaging, big data bridge these levels of organization. This review explores the challenges opportunities in linking microscopic phenomena macroscopic functions, emphasizing methodologies driving progress field. It also highlights clinical potential including their role advancing artificial intelligence (AI) applications improving healthcare technologies. By examining current research proposing future directions for interdisciplinary collaboration, work demonstrates how can revolutionize both scientific understanding practice.

Язык: Английский

---

Developing a brain inspired multilobar neural networks architecture for rapidly and accurately estimating concrete compressive strength

Scientific Reports, Год журнала: 2025, Номер 15(1)

Опубликована: Янв. 15, 2025

Concrete compressive strength is a critical parameter in construction and structural engineering. Destructive experimental methods that offer reliable approach to obtaining this property involve time-consuming procedures. Recent advancements artificial neural networks (ANNs) have shown promise simplifying task by estimating it with high accuracy. Nevertheless, conventional ANNs often require deep achieve acceptable results cases large datasets where generalization required for variety of mixtures. This leads increased training durations susceptibility noise, causing reduced accuracy potential information loss networks. In order address these limitations, study introduces novel multi-lobar network (MLANN) architecture inspired the brain's lobar processing sensory information, aiming improve efficiency concrete strength. The MLANN framework employs various architectures hidden layers, referred as "lobes," each unique arrangement neurons optimize data processing, reduce expedite time. Within context, an developed, its performance evaluated predict concrete. Moreover, compared against two traditional cases, ANN ensemble learning (ELNN). indicated significantly improves estimation performance, reducing root mean square error up 32.9% absolute 25.9% while also enhancing A20 index 17.9%, ensuring more robust generalizable model. advancement model refinement can ultimately enhance design analysis processes civil engineering, leading cost-effective practices.

Язык: Английский

---

To Spike or Not to Spike: A Digital Hardware Perspective on Deep Learning Acceleration

IEEE Journal on Emerging and Selected Topics in Circuits and Systems, Год журнала: 2023, Номер 13(4), С. 1015 - 1025

Опубликована: Ноя. 6, 2023

As deep learning models scale, they become increasingly competitive from domains spanning computer vision to natural language processing; however, this happens at the expense of efficiency since require more memory and computing power. The power biological brain outperforms any large-scale (DL) model; thus, neuromorphic tries mimic operations, such as spike-based information processing, improve DL models. Despite benefits brain, efficient transmission, dense neuronal interconnects, co-location computation memory, available substrate has severely constrained evolution brains. Electronic hardware does not have same constraints; therefore, while modeling spiking neural networks (SNNs) might uncover one piece puzzle, design backends for SNNs needs further investigation, potentially taking inspiration work done on artificial (ANNs) side. such, when is it wise look designing new hardware, should be ignored? To answer question, we quantitatively compare digital acceleration techniques platforms ANNs SNNs. a result, provide following insights: (i) currently process static data efficiently, (ii) applications targeting produced by sensors, event-based cameras silicon cochleas, need investigation behavior these sensors naturally fit SNN paradigm, (iii) hybrid approaches combining lead best solutions investigated level, accounting both loss optimization.

Язык: Английский

---

AI-Enhanced ECG Applications in Cardiology: Comprehensive Insights from the Current Literature with a Focus on COVID-19 and Multiple Cardiovascular Conditions

Diagnostics, Год журнала: 2024, Номер 14(17), С. 1839 - 1839

Опубликована: Авг. 23, 2024

The application of artificial intelligence (AI) in electrocardiography is revolutionizing cardiology and providing essential insights into the consequences COVID-19 pandemic. This comprehensive review explores AI-enhanced ECG (AI-ECG) applications risk prediction diagnosis heart diseases, with a dedicated chapter on COVID-19-related complications. Introductory concepts AI machine learning (ML) are explained to provide foundational understanding for those seeking knowledge, supported by examples from literature current practices. We analyze ML methods arrhythmias, failure, pulmonary hypertension, mortality prediction, cardiomyopathy, mitral regurgitation, embolism, myocardial infarction, comparing their effectiveness both medical perspectives. Special emphasis placed cardiology, including detailed comparisons different methods, identifying most suitable approaches specific analyzing strengths, weaknesses, accuracy, clinical relevance, key findings. Additionally, we explore AI's role emerging field cardio-oncology, particularly managing chemotherapy-induced cardiotoxicity detecting cardiac masses. serves as an insightful guide call action further research collaboration integration aiming enhance precision medicine optimize decision-making.

Язык: Английский

---

Adoption of Deep-Learning Models for Managing Threat in API Calls with Transparency Obligation Practice for Overall Resilience

Sensors, Год журнала: 2024, Номер 24(15), С. 4859 - 4859

Опубликована: Июль 26, 2024

System-to-system communication via Application Programming Interfaces (APIs) plays a pivotal role in the seamless interaction among software applications and systems for efficient automated service delivery. APIs facilitate exchange of data functionalities across diverse platforms, enhancing operational efficiency user experience. However, this also introduces potential vulnerabilities that attackers can exploit to compromise system security, highlighting importance identifying mitigating associated security risks. By examining weaknesses inherent these using open-intelligence catalogues like CWE CAPEC implementing controls from NIST SP 800-53, organizations significantly enhance their posture, safeguarding against threats. task is challenging due evolving threats vulnerabilities. Additionally, it analyse given large volume traffic generated API calls. This work contributes tackling challenge makes novel contribution managing within system-to-system through It an integrated architecture combines deep-learning models, i.e., ANN MLP, effective threat detection call datasets. The identified are analysed determine suitable mitigations improving overall resilience. Furthermore, transparency obligation practices entire AI life cycle, dataset preprocessing model performance evaluation, including methodological SHapley Additive exPlanations (SHAP) analysis, so models understandable by all groups. proposed methodology was validated experiment Windows PE Malware dataset, achieving average accuracy 88%. outcomes experiments summarized provide list key features, such as FindResourceExA NtClose, which linked with related threats, order identify accurate control actions manage

Язык: Английский

---

Optimizing and investigating the charging time of phase change materials in a compact-latent heat storage using pareto front analysis, artificial neural networks, and numerical simulations

Journal of Energy Storage, Год журнала: 2024, Номер 102, С. 113966 - 113966

Опубликована: Окт. 5, 2024

Язык: Английский

---