English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

A fundamental question for understanding brain function is what types ofstimuli drive neurons to fire. In visual neuroscience, this question has alsobeen posted as characterizing the receptive field of a neuron. The search foreffective stimuli has traditionally been based on a combination of insightsfrom previous studies, intuition, and luck. Recently, the same question hasemerged in the study of units in convolutional neural networks (ConvNets), andtogether with this question a family of solutions were developed that aregenerally referred to as "feature visualization by activation maximization."  We sought to bring in tools and techniques developed for studying ConvNets tothe study of biological neural networks. However, one key difference thatimpedes direct translation of tools is that gradients can be obtained fromConvNets using backpropagation, but such gradients are not available from thebrain. To circumvent this problem, we developed a method for gradient-freeactivation maximization by combining a generative neural network with a geneticalgorithm. We termed this method XDream (EXtending DeepDream with real-timeevolution for activation maximization), and we have shown that this method canreliably create strong stimuli for neurons in the macaque visual cortex (Ponceet al., 2019). In this paper, we describe extensive experiments characterizingthe XDream method by using ConvNet units as in silico models of neurons. Weshow that XDream is applicable across network layers, architectures, andtraining sets; examine design choices in the algorithm; and provide practicalguides for choosing hyperparameters in the algorithm. XDream is an efficientalgorithm for uncovering neuronal tuning preferences in black-box networksusing a vast and diverse stimulus space.

Gradient-free activation maximization for identifying effective stimuli