A neural network model is proposed to achieve invariant pattern recognition to binary inputs based on a combination of the invariance net and the ART1 neural network. Computer simulations show that the new network is plastic to unfamiliar inputs and requires no explicit training of the inputs.