Speech recognition with spiking neurons and dynamic synapses: a model motivated by the human auditory pathway

Näger, Christian; Storck, Jan; Deco, Gustavo

doi:10.1016/s0925-2312(02)00494-0

Cited by 17 publications

(10 citation statements)

References 4 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…This precise temporal pattern in spiking activity is considered as a crucial coding strategy in sensory information processing areas [20], [21], [22], [23], [24] and neural motor control areas in the brain [25], [26]. SNNs have become the focus of a number of recent applications in many areas of pattern recognition such as visual processing [27], [28], [29], [30], speech recognition [31], [32], [33], [34], [35], [36], [37], and medical diagnosis [38], [39]. In recent years, a new generation of neural networks that incorporates the multilayer structure of DNNs (and the brain) and the type of information communication in SNNs has emerged.…”

Section: Introductionmentioning

confidence: 99%

Deep learning in spiking neural networks

et al. 2019

View full text Add to dashboard Cite

1 In recent years, deep learning has revolutionized the field of machine learning, for computer vision in particular. In this approach, a deep (multilayer) artificial neural network (ANN) is trained in a supervised manner using backpropagation. Vast amounts of labeled training examples are required, but the resulting classification accuracy is truly impressive, sometimes outperforming humans. Neurons in an ANN are characterized by a single, static, continuous-valued activation. Yet biological neurons use discrete spikes to compute and transmit information, and the spike times, in addition to the spike rates, matter. Spiking neural networks (SNNs) are thus more biologically realistic than ANNs, and arguably the only viable option if one wants to understand how the brain computes. SNNs are also more hardware friendly and energy-efficient than ANNs, and are thus appealing for technology, especially for portable devices. However, training deep SNNs remains a challenge. Spiking neurons' transfer function is usually non-differentiable, which prevents using backpropagation.Here we review recent supervised and unsupervised methods to train deep SNNs, and compare them in terms of accuracy, but also computational cost and hardware friendliness. The emerging picture is that SNNs still lag behind ANNs in terms of accuracy, but the gap is decreasing, and can even vanish on some tasks, while SNNs typically require many fewer operations.

show abstract

Section: Introductionmentioning

confidence: 99%

Deep learning in spiking neural networks

et al. 2019

View full text Add to dashboard Cite

show abstract

“…Recently, several models have also attempted to reproduce, to some extent, the structure of the visual cortex [64,15,47,55], as well as an earlier version of this model [43], on which our work is based. Not many neural models have been proposed for the auditory process [39,61], and little is yet known about the kind of brain computations that lead to word recognition there. Recently, [27] addressed the important issue of the roles and the interactions between ITC (Inferior Temporal Cortex) and PFC (Pre-Frontal Cortex) in categorization, with a neural model, which was again limited to vision without any relation to words.…”

Section: Computational Approachesmentioning

confidence: 99%

A Biologically Inspired Neural Model of Vision-Language Integration

Plebe

Mazzone

Cruz

2011

NNW

View full text Add to dashboard Cite

Nowadays, remote sensing technology is being used as an essential tool for monitoring and detecting oil spills to take precautions and to prevent the damages to the marine environment. As an important branch of remote sensing, satellite based synthetic aperture radar imagery (SAR) is the most effective way to accomplish these tasks. Since a marine surface with oil spill seems as a dark object because of much lower backscattered energy, the main problem is to recognize and differentiate the dark objects of oil spills from others to be formed by oceanographic and atmospheric conditions. In this study, Radarsat-1 images covering Lebanese coasts were employed for oil spill detection. For this purpose, a powerful classifier, Artificial Neural Network Multilayer Perceptron (ANN MLP) was used. As the original contribution of the paper, the network was trained by a novel heuristic optimization algorithm known as Artificial Bee Colony (ABC) method besides the conventional Backpropagation (BP) and Levenberg-Marquardt (LM) learning algorithms. A comparison and evaluation of different network training algorithms regarding reliability of detection and robustness show that for this problem best result is achieved with the Artificial Bee Colony algorithm (ABC).

show abstract

“…Les poids k des canaux en ordre croissant, de 1 à 24, sont donc : 0, 0, 1, 0, 0, 3,4,2,7,6,9,14,13,15,20,17,19,18,16,11,12 En observant les résultats de la reconnaissance (tableau A.4), il est surprenant de voir qu'un grand pourcentage des prononciations est classé correctement (79 %). De plus, les résultats obtenus avec le cinquième locuteur (mpe), non utilisés pour la création des modèles, sont de l'ordre de 90 %.…”

Section: A 15 Exempleunclassified

Exploration de réseaux de neurones à décharges dans un contexte de reconnaissance de parole /

Loiselle¹

2004

View full text Add to dashboard Cite

Speech recognition with spiking neurons and dynamic synapses: a model motivated by the human auditory pathway

Cited by 17 publications

References 4 publications

Deep learning in spiking neural networks

Deep learning in spiking neural networks

A Biologically Inspired Neural Model of Vision-Language Integration

Exploration de réseaux de neurones à décharges dans un contexte de reconnaissance de parole /

Contact Info

Product

Resources

About