Active sensing with artificial neural networks

The fitness of behaving agents depends on their knowledge the environment, which demands efficient exploration strategies. Active sensing formalizes as reduction uncertainty about current state environment. Despite strong theoretical justifications, active has had limited applicability due to difficulty in estimating information gain. Here we address this issue by proposing a linear approximation gain and implementing gradient-based action selection within an artificial neural network setting. We compare estimation with art, validate our model task based MNIST dataset. also propose that exploits amortized inference network, performs equally well certain contexts.