We propose an Artificial Neural Net (ANN) architecture for discovering common hidden variables and for learning of invariant representations through synchronicity, coincidence and concurrence. In the common variable discovery problem, the ANN uses measurements from two distinct sensors to construct a representation of the common hidden variable that is manifested in both sensors, and discards sensor-specific variables. In the invariant representation learning problem, the network uses multiple observations of objects under transformations to construct a representation which is invariant to the transformations. Unlike classic regression problems, the network is not presented with labels to learn, instead, it must infer a proper representation form the data; unlike classic signal processing problems, the algorithm is not given the invariant features to compute, and it must discover proper features. Common Variable Discovery and Invariant Representation Learning using Artificial Neural Networks Uri Shaham † and Roy R. Lederman ‡ , Technical Report YALEU/DCS/TR-1506 † Department of Statistics, Yale University, New Haven CT 06511 ‡ Applied Mathematics Program, Yale University, New Haven CT 06511 Approved for public release: distribution is unlimited. Keywords: common variable, hidden variable, invariance learning, equivalence learning, rep- resentation learning, coincidence modeling, neural networks