Pre-activation Distributions Expose Backdoor Neurons.

Runkai Zheng Rongjun Tang Jianze Li Li Liu

Published in: NeurIPS (2022)

Keyphrases

neural network
probability distribution
lateral inhibition
information processing
associative memory
kullback leibler divergence
random variables
receptive fields
power law
single neuron
natural images
high dimensional
hidden layer
probability density
exponential family
spiking neurons
data sets