Deep recurrent attentional neural network model simulating human mathematical cognition. Demonstrating human-like development of ability to estimate and count number.
A fully differentiable recurrent visual attention model to count the number of blobs. Some code / ideas taken from this implementation (https://github.com/ericjang/draw) of this paper (http://arxiv.org/pdf/1502.04623.pdf), and built upon this implementation (https://github.com/jlindsey15/DRAM).
Deep recurrent attentional neural network model simulating human mathematical cognition. Demonstrating human-like development of ability to estimate and count number.