Biometrics is the automatic recognition of people using their physical appearance, behavioral traits, and/or their compound effects. Among biometric modalities (fingerprints, iris, hand geometry, etc.), face recognition has gained most interest due to its low cost, passive mode of operation, and wide acceptance. However, face-based biometric systems must overcome significant challenges when deployed in uncontrolled settings due to the variability of facial appearance. In this thesis, we focus on the variability due to face aging and biometric interoperability. Towards that end we present novel approaches to face recognition subject to time lapse with emphasis on deep learning and automatic rather handcrafted feature extraction, demographic stratification, and set based metrics for similarity biometric matching.