Digital selection and analogue amplification coexist in a cortex-inspired silicon circuit
By Richard Hans Robert Hahnloser, Rahul Sarpeshkar, M. Mahowald, Rodney J. Douglas, H. Sebastian Seung. 2000
1,317 citations recorded by the source corpus.
Subfields: Artificial Intelligence, Cognitive Neuroscience, Electrical and Electronic Engineering
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Sergey Levine
- Yoshua Bengio
- Quoc V. Le
- Pieter Abbeel
- Mohit Bansal
- Ruslan Salakhutdinov
- Jianfeng Gao
- Dacheng Tao
In Cognitive Neuroscience
- Charles F. Cadieu
- Wolfgang Maass
- Tomaso A. Poggio
- Minjoon Kouh
- Thomas Serre
- James J. DiCarlo
- Timothy P. Lillicrap
- Nikolaus Kriegeskorte
In Electrical and Electronic Engineering
- Wolfgang Maass
- Timothy P. Lillicrap
- Michael Pfeiffer
- Demis Hassabis
- David Silver
- Quoc V. Le
- Lars Buesing
- Diganta Misra
Citations in the corpus
- Medical Image Analysis using Convolutional Neural Networks: A Review (2017, background): “This is due to the availability of improved computational capabilities, nonlinearities which allow for deeper networks [73, 74], and better ways to initialize the deep network [75].”
- Privacy-Preserving Classification on Deep Neural Network (2017, methodology): “the introduction of the dropout regularization technique [27] and the introduction of the REctified Linear Unit activation function [24]).”
- Bayesian Computation Emerges in Generic Cortical Microcircuits through Spike-Timing-Dependent Plasticity (2013, background): “These ubiquitous network motifs were called soft winner-take-all (WTA) circuits, and have been suggested as neural network models for implementing functions like non-linear selection [16,17], normalization [18], selective attention [19], decision making [20,21], or as primitives for general purpose ”
- Improving Deep Learning using Generic Data Augmentation (2017, methodology): “All neurons used a Rectified Linear Unit [Hahnloser et al., 2000] with the weights initially being initialised from a Gaussian distribution with a 0 mean and a standard deviation of 0.01.”
- Deep Learning using Rectified Linear Units (ReLU) (2018, background): “ently, we have pk = exp(ok) ˝ n−1 k=0 exp(ok) (3) Hence, the predicted class would be yˆ yˆ = argmax i∈1,...,N pi (4) 2.4.2 Rectified Linear Units (ReLU). ReLU is an activation function introduced by [6], which has strong biological and mathematical underpinning. In 2011, it was demonstrated to furt”
- Deep learning for determining a near-optimal topological design without any iteration (2018, background): “The advent of a new activation function called the rectified linear unit (ReLU) (Hahnioser et al. 2000; Nair and Hinton 2010) and effective weight initialization (Hinton et al. 2006) allowed for rapidly processing a considerable amount of data using enhanced hardware such as general-purpose…”