In recent years, deep learning has significantly improved the fields of computer vision, natural language processing and speech recognition. Beyond these traditional fields, deep learning has been expended to quantum chemistry, physics, neuroscience, and more recently to combinatorial optimization (CO). Well-known CO problems are Travelling Salesman Problem, assignment problems, routing, planning, Bayesian search, and scheduling. CO is basically used every day in finance and revenue management, transportation, manufacturing, supply chain, public policy, hardware design, computing and information technology. The workshop will bring together experts in mathematics (optimization, graph theory, sparsity, combinatorics, statistics), CO (assignment problems, routing, planning, Bayesian search, scheduling), machine learning (deep learning, supervised, self-supervised and reinforcement learning) and specific applicative domains (e.g. finance, transportation, hardware design, computing and information technology) to establish the current state of these emerging techniques and discuss the next directions. Besides, such generalization of deep learning techniques to CO problems will also push forward the mathematical analysis of the properties of these learning systems like generalization and transfer, stochastic optimization and dynamic predictivity that make the success of these techniques.