Energy-efficient neural network chips approach human recognition capabilities

Maass, Wolfgang

doi:10.1073/pnas.1614109113

Cited by 13 publications

(15 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Spiking networks offer an alternative solution by exploiting event-based data-driven computations that makes them attractive for deployment on real-time neuromorphic hardware where power consumption and speed become vital constraints (Merolla et al, 2014 ; Han et al, 2015 ). While research efforts in spiking models have soared in the recent past (Masquelier and Thorpe, 2007 ; Srinivasa and Cho, 2012 ; Maass, 2016 ; Panda and Roy, 2016 ), the performance of such models are not as accurate as compared to their artificial counterparts (or DLNs). From our perspective, this work provides a new standard establishing the effectiveness of spiking models and their inherent timing-based processing for action recognition in videos.…”

Section: Resultsmentioning

confidence: 99%

Learning to Recognize Actions From Limited Training Examples Using a Recurrent Spiking Neural Model

Panda

Srinivasa

2018

Front. Neurosci.

View full text Add to dashboard Cite

A fundamental challenge in machine learning today is to build a model that can learn from few examples. Here, we describe a reservoir based spiking neural model for learning to recognize actions with a limited number of labeled videos. First, we propose a novel encoding, inspired by how microsaccades influence visual perception, to extract spike information from raw video data while preserving the temporal correlation across different frames. Using this encoding, we show that the reservoir generalizes its rich dynamical activity toward signature action/movements enabling it to learn from few training examples. We evaluate our approach on the UCF-101 dataset. Our experiments demonstrate that our proposed reservoir achieves 81.3/87% Top-1/Top-5 accuracy, respectively, on the 101-class data while requiring just 8 video examples per class for training. Our results establish a new benchmark for action recognition from limited video examples for spiking neural models while yielding competitive accuracy with respect to state-of-the-art non-spiking neural models.

show abstract

Section: Resultsmentioning

confidence: 99%

Learning to Recognize Actions From Limited Training Examples Using a Recurrent Spiking Neural Model

Panda

Srinivasa

2018

Front. Neurosci.

View full text Add to dashboard Cite

show abstract

“…This is now beginning to change by the recent technological development (IBM, TrueNorth ), making it possible to emulate spiking neural networks directly in state‐of‐the‐art semiconductor hardware, building extremely fast event‐based systems for image processing and deep learning …”

Section: What Is the Power Of Computing?mentioning

confidence: 99%

Can Biological Quantum Networks Solve NP‐Hard Problems?

Wendin

2019

Adv Quantum Tech

View full text Add to dashboard Cite

There is a widespread view that the human brain is so complex that it cannot be efficiently simulated by universal Turing machines, let alone ordinary classical computers. During the last decades the question has therefore been raised whether it is needed to consider quantum effects to explain the imagined cognitive power of a conscious mind. Not surprisingly, the conclusion is that quantum‐enhanced cognition and intelligence are very unlikely to be found in biological brains. Quantum effects may certainly influence signaling pathways at the molecular level in the brain network, like ion ports, synapses, sensors, and enzymes. This might evidently influence the functionality of some nodes and perhaps even the overall intelligence of the brain network, but hardly give it any dramatically enhanced functionality. The conclusion is that biological quantum networks can only approximately solve small instances of nonpolynomial (NP)‐hard problems. On the other hand, artificial intelligence and machine learning implemented in complex dynamical systems based on genuine quantum networks can certainly be expected to show enhanced performance and quantum advantage compared with classical networks. Nevertheless, even quantum networks can only be expected to solve NP‐hard problems approximately. In the end it is a question of precision—Nature is approximate.

show abstract

“…As for the realizations, even if the networks have an autonomous structure, generally computer simulations are used in order to allow even substantial modifications in a short time and with limited costs. However, the first neural chips [12] are being created that have a performance considerably higher than that of a simulation but that has so far had very little diffusion due mainly to high costs and extreme structural rigidity.…”

Section: Artificial Neural Network For Supervised Learningmentioning

confidence: 99%

“…is the stabilizing term that makes the sum of the (12) limited and close to 1 without explicit normalisation appearing. The Oja rule can be generalized for networks that have multiple output neurons by obtaining the two algorithms in Figure 5 and Figure 6.…”

Section: Neural Implementation Of Standard Pcamentioning

confidence: 99%

Feature Selection with Non Linear PCA: A Neural Network Approach

Gallo¹,

Capozzi²

2019

JAMP

View full text Add to dashboard Cite

Machine learning consists in the creation and development of algorithms that allow a machine to learn itself, gradually improving its behavior over time. This learning is more effective, the more representative is the features of the dataset used to describe the problem. An important objective is therefore the correct selection (and, possibly, reduction of the number) of the most relevant features, which is typically carried out through dimensional reduction tools such as Principal Component Analysis (PCA), which is not linear in the more general case. In this work, an approach to the calculation of the reduced space of the PCA is proposed through the definition and implementation of appropriate models of artificial neural network, which allows to obtain an accurate and at the same time flexible reduction of the dimensionality of the problem.

show abstract

Energy-efficient neural network chips approach human recognition capabilities

Cited by 13 publications

References 10 publications

Learning to Recognize Actions From Limited Training Examples Using a Recurrent Spiking Neural Model

Learning to Recognize Actions From Limited Training Examples Using a Recurrent Spiking Neural Model

Can Biological Quantum Networks Solve NP‐Hard Problems?

Feature Selection with Non Linear PCA: A Neural Network Approach

Contact Info

Product

Resources

About