As the research interest in oceans and seas increases, autonomous submarines are becoming a viable research tool. The study of autonomous navigation has flourished in recent decades. However, underwater vehicles pose unique challenges. Tools and methodologies that are applicable for air and ground navigation (e.g., GPS) are, in many cases, irrelevant for underwater vehicles. Traditional approaches for submarine localization suffer from deteriorated performance as the noise level increases. In this work, we propose a data-driven approach to localize submarines. Using the Deep Learning technique, we predict the submarine’s location from acoustic sensor data. We evaluated our approach in simulation and demonstrated improved accuracy compared with a traditional Iterative Least Squares algorithm.