The design of new functional materials has been a critical part of emerging technologies over the past century. Advancements in energy storage, drug delivery, solar energy, filtration, carbon capture and semiconductors have accelerated due to the discovery of entire classes of materials with application specific properties, such as Perovskites and metal-organic frameworks (MOFs). However, ab initio computational methods [2] for designing new inorganic materials are slow and scale poorly to realistically sized systems. New methods using deep learning offer a way to achieve ab initio accuracy with dramatically increased speed and scalability. In recent years, deep learning methods have demonstrated their ability to approximate extremely complex natural distributions across a diverse range of application areas including vision, biology and spatial processing, by focusing on architectures that are embarrassingly parallel and can be run efficiently on modern hardware [46, 7], despite lacking architectural biases which would suit the target domain.

The exploration of thermoelectric materials is challenging considering the large materials space, combined with added exponential degrees of freedom coming from doping and the diversity of synthetic pathways. Here we seek to incorporate historical data and update and refine it using experimental feedback by employing error-correction learning (ECL). We thus learn from prior datasets and then adapt the model to differences in synthesis and characterization that are otherwise difficult to parameterize. We then apply this strategy to discovering thermoelectric materials where we prioritize synthesis at temperatures < 300{\deg}C. We document a previously unreported chemical family of thermoelectric materials, PbSe:SnSb, finding that the best candidate in this chemical family, 2 wt% SnSb doped PbSe, exhibits a power factor more than 2x that of PbSe. Our investigations show that our closed-loop experimentation strategy reduces the required number of experiments to find an optimized material by as much as 3x compared to high-throughput searches powered by state-of-the-art machine learning models. We also observe that this improvement is dependent on the accuracy of prior in a manner that exhibits diminishing returns, and after a certain accuracy is reached, it is factors associated with experimental pathways that dictate the trends.