Learning Without Feedback: Fixed Random Learning Signals Allow for Feedforward Training of Deep Neural Networks

While the backpropagation of error algorithm enables deep neural network training, it implies (i) bidirectional synaptic weight transport and (ii) update locking until the forward and backward passes are completed. Not only do these constraints preclude biological plausibility, but they also hinder the development of low-cost adaptive smart sensors at the edge, as they severely constrain memory accesses and entail buffering overhead. In this work, we show that the one-hot-encoded labels provided in supervised classification problems, denoted as targets, can be viewed as a proxy for the error sign. Therefore, their fixed random projections enable a layerwise feedforward training of the hidden layers, thus solving the weight transport and update locking problems while relaxing the computational and memory requirements. Based on these observations, we propose the direct random target projection (DRTP) algorithm and demonstrate that it provides a tradeoff between accuracy and computational cost that is suitable for adaptive edge computing devices.

Disrupting flight increases sleep and identifies a novel sleep-promoting pathway in Drosophila

Sleep is plastic and is influenced by ecological factors and environmental changes. The mechanisms underlying sleep plasticity are not well understood. We show that manipulations that impair flight in Drosophila increase sleep as a form of sleep plasticity. We disrupted flight by blocking the wing-expansion program, genetically disrupting flight, and by mechanical wing perturbations. We defined a new sleep regulatory circuit starting with specific wing sensory neurons, their target projection neurons in the ventral nerve cord, and the neurons they connect to in the central brain. In addition, we identified a critical neuropeptide (burs) and its receptor (rickets) that link wing expansion and sleep. Disrupting flight activates these sleep-promoting projection neurons, as indicated by increased cytosolic calcium levels, and stably increases the number of synapses in their axonal projections. These data reveal an unexpected role for flight in regulating sleep and provide new insight into how sensory processing controls sleep need.

Morphisms which are continuous on a neighborhood of the base of a groupoid

Kirill Mackenzie raised in [3] (p. 31) the following question: given a morphism F : Ω → Ω′, where Ω and Ω′ are topological groupoids and F is continuous on a neighborhood of the base in Ω, is it true that is Ω continuous everywhere?This paper gives a negative answer to that question. Moreover, we shall prove that for a locally compact groupoid Ω with non-singleton orbits and having open target projection, if we assume that the continuity of every morphism F on a neighborhood of the base in Ω implies the continuity of F everywhere, then the groupoid Ω must be locally transitive.