An introduction to
Alan Dix with Janet Finlay
These are some of the topics I’m considering for the new edition, but this may change as things develop – so please let me know what you think!
deep learning – multi-level NN, reinforcement learning, adversarial learning
dimension reduction – classic factorial and similarity matrix decompositions, restricted Boltzmann machines, random vectors
explainable AI – argumentation, deep learning and black-box ML, sensitivity/perturbation analysis
human-centred AI – bias (gender, ethnic), need for transparency, augmentation vs replacement, IUI, designing interactions to aid ML
big data – recommenders, surprising power (e.g. language), map-reduce, limits and links to HPC, large network analysis
in the world – more robots, autonomous vehicles, smart homes, smart cities
temporal AI – recurrent NN, HMM, applications: language, activity recognition, security, continuous data (e.g. GAs for ODEs)
human-like computing – one shot learning, higher-level AI, links to explainability
practicalities – data cleaning, reconciliation, mixed methods
plus – spiking NN, algebraic topology and other new techniques, ontologies and sem-web for knowledge representation, update ethics/legal/philosophical
case studies – small (1-2 page) boxes or in text – maybe over time add longer case studies to web: onCue, GA for submarine design, musicology data, AlphaGo, Watson, QbB, find security example, predictive policing?