Interested in getting a 90-hour certificate in Autonomous Systems? Then register using this page, which bundles together all three of the courses in the Introduction to Mission Critical Autonomous Artificial Intelligence Systems for easy registration and check out. The courses are as follows:
In this course, students will be introduced to Probabilistic Graph Models and Probabilistic Programming while learning about and implementing Bayesian estimation of dynamical systems. In addition, the student will apply machine learning for pattern recognition, prediction, and sequential updating of knowledge captured by the programs.
In this course, students will be introduced to the key elements of machine learning, namely: Credit Assignment Problems and Regularization. Assigning credit, or blame, is to calculate how model parameters should be estimated, or updated, to fit the training data. Regularization smooths and tempers these updates to avoiding overfitting the training data and promote generalization when presented with new data. Of course, the latter is the whole point. Stochastic processes and optimization come into play much like how a smith tempers steel. The student will learn how and why machine learning is essentially stochastic optimization where randomness can work with, rather than against, us. In so doing, the student will learn to apply probabilistic programming and stochastic optimization methods to latent, state-based model estimation. The student will also learn to develop learning algorithms for sequence data and apply the generated programs to applications that involve time series predictions and sequential decision making.
In this course, students will learn how to use probabilistic programming to build models and supporting algorithms that learn sequentially under uncertainty. The goal is to learn a Markov decision process that informs the program on how to optimally interact with its real-world environment in real-time. The student will learn how this underlying stochastic process is constructed with latent-variable graph models that enjoy exact or approximate numerical solutions. The type of solution depends on certain properties a given graph possesses. The student will learn how to use a high-level language for probabilistic programming that emits coded graph model descriptions accompanied with the appropriate learning and inference algorithm. We call this innovation the “model of computation” and executing the generated program performs the exact/approximate marginalization to compute answers to statistical questions. Students will learn how to apply these tools and machinery to engineer AI systems and becomes the basis for online learning and operations with streaming data. The student will take away an understanding of how such systems learn to perform sequential tasks, improve incrementally, and interact with the world online in real-time. Moreover, this formal, model-based framework introduced in this course has application to machine learning engineering that is principled, transparent, and hence, verifiable for safety-critical systems.
If you would like to enroll for all courses at once, please visit: AUTO.PRO1