Compute & Intelligence
Features, embeddings, training regimes and evaluation for sensor data.
A Nature Reviews survey of computational light-field detection, alongside a separately funded Rice project on generative cameras, points at imaging systems that measure less and reconstruct more.
A Yale-led team used machine learning to predict how similar two complex odour mixtures smell to people — the clearest step yet toward quantitative olfaction, and still a long way from an artificial nose.
One diagram, seven layers. From a physical phenomenon through the analog front end, converter, preprocessing, features and embeddings, into a model and finally a decision that moves an actuator — and the failure mode that lives in each layer.