Deep neural networks using a single neuron: folded-in-time architecture using feedback-modulated delay loops
Stelzer, Florian; Röhm, Andre; Vicente, Raul; Fischer, Ingo; Yanchuk, Serhiy
Nature Communications 12, 5164 (1-10) (2021)
Nature Communications 12, 5164 (1-10) (2021)
Deep neural networks are among the most widely applied machine learning tools showing outstanding performance in a broad range of tasks. We present a method for folding a deep neural network of arbitrary size into a single neuron with multiple time-delayed feedback loops. This single-neuron deep neural network comprises only a single nonlinearity and appropriately adjusted modulations of the feedback signals. The network states emerge in time as a temporal unfolding of the neuron’s dynamics. By adjusting the feedback-modulation within the loops, we adapt the network’s connection weights. These connection weights are determined via a back-propagation algorithm, where both the delay-induced and local network connections must be taken into account. Our approach can fully represent standard Deep Neural Networks (DNN), encompasses sparse DNNs, and extends the DNN concept toward dynamical systems implementations. The new method, which we call Folded-in-time DNN (Fit-DNN), exhibits promising performance in a set of benchmark tasks.
Press release
Additional files
Related research projects
Noticias relacionadas
Deep Neural Networks using a single neuron
14 de septiembre de 2021
Deep Neural Networks (DNNs) are a useful tool for a wide range of tasks such as image classification, object detection, image resizing or text generation. They are extremely powerful, but they require sufficient computing power and large data sets to …
Xarxes Neuronals Profundes amb només una neurona
14 de septiembre de 2021
Les xarxes neuronals profundes (DNN) són una eina útil per a una àmplia gamma de tasques, com la classificació d'imatges, la detecció d'objectes, el redimensionament d'imatges o la generació de textos. Són extremadament potents, però requereixen una bona capacitat de …
Redes Neuronales Profundas con solo una neurona
14 de septiembre de 2021
Las redes neuronales profundas (DNN) son una herramienta útil para una amplia gama de tareas como la clasificación de imágenes, la detección de objetos, el redimensionamiento de imágenes o la generación de textos. Son extremadamente potentes, pero requieren suficiente capacidad …