research mothership & drone routing An exact branch-and-price-and-cut algorithm for coordinating a support vehicle with a fleet of delivery drones. learning to price Deep reinforcement learning for prioritizing pricing subproblems inside column generation. learning cutting planes Deep Q-learning for generating Chvátal–Gomory cuts, plus a hybrid inequality family. network resilience after snow storms Measuring and forecasting how New York City's transportation network recovers from winter storms. project networks with graph neural networks Predicting project duration and cost from activity–resource graphs, beyond CPM and PERT. applied ml stock price forecasting with LSTM An LSTM trained on eight years of AAPL data, judged on direction rather than on loss. CNN hyperparameter search on CIFAR-10 Optuna-driven tuning of a convolutional network, optimising for search efficiency rather than raw accuracy. English-to-French translation with T5 Fine-tuning T5-small on OPUS Books with mixed-precision training. Q-learning for shortest paths Reinforcement learning on a constrained grid graph, as a bridge between RL and combinatorial optimization.
