# FYS4447/9447 – Advanced machine learning and data analysis for the physical sciences ### Level: Master and PhD ### Credits: 10 ### Teaching: spring semester ### Examination: Every Spring ### Teaching language: Norwegian (English on request) ## Course content Advances in artificial intelligence/machine learning methods provide tools that have broad applicability in scientific research. These techniques are being applied across the diversity of research topics in modern science, leading to advances that will facilitate scientific discoveries and societal applications. This course focuses on advanced machine learning and statistical learning methods applied to a broad variety of problems in the physical sciences and life science, from computational neuroscience to the analysis of high-energy physics experiments. Supervised and unsupervised learning methods are discussed, spanning from various deep learning methods to Bayesian modeling. After this course you should : - be familiar with central deep learning methods and how to use them in actual research - be familiar with advanced regression algorithms - understand how to to simulate complex physical processes with many degrees of freedom - understand optimization techniques and their fundamental role in machine learning - be familiar with Bayesian statistics and Bayesian Machine Learning - understand how to find correlations in data sets and quantify uncertainties - understand how to use Gaussian processes in the analysis of physics problems ## Admission Students admitted at UiO must apply for courses in Studentweb. Students enrolled in other Master's Degree Programmes can, on application, be admitted to the course if this is cleared by their own study programme. Nordic citizens and applicants residing in the Nordic countries may apply to take this course as a single course student. If you are not already enrolled as a student at UiO, please see our information about admission requirements and procedures for international applicants. ## Prerequisites Recommended previous knowledge A good background in mathematics is needed. Other recommended courses are: FYS-STK4155 – Applied Data Analysis and Machine Learning IN5400 - Machine learning for image analysis ## Overlapping courses None ## Teaching The course is based on assigned self studies and three projects. There are no regular lectures. Meetings every week (lasting two hours) are arranged in order to monitor progress and discuss projects and exercises. ## Examination Three project assignments that each is given a weight of 1/3 in the final grading (100% together). There is no final exam. ### Examination support material No examination support material is allowed. ## Language of examination You may write your examination papers in Norwegian, Swedish, Danish or English. ## Grading scale Grades are awarded on a scale from A to F, where A is the best grade and F is a fail. Read more about the grading system. For PhD students it is only pass/not passed. ## Detailed content (this may vary according to the interests of the participants, but what is listed here covers the main topics). These topics will be tailored to actual data from experiments and/or simulations of physical systems. ### Advanced deep learning methods - The structure of convolutional and recurrent neural networks - High-level concepts in deep neural networks - Organizing deep learning workflows using the bias–variance tradeoff - Deep generative models: Hidden variables and restricted Boltzmann machines (RBMs) - Autoencoders, Variational autoencoders and generative adversarial networks ### Dimensional reduction and data visualization - Principal component analysis (PCA) - Multidimensional scaling - Clustering and k-means - Hierarchical clustering: Agglomerative methods ### Bayesian Optimization - Basics of Bayesian inference - Fundamentals of Bayesian data analysis - Bayesian computations - Markov chain Monte Carlo simulations, Gibbs and Metropolis sampling - Bayesian Regression models - Bayesian machine learning - Gaussian process models