Learning Binary Latent Variable Models: A Tensor Eigenpair Approach