In this class, students will learn about the theoretical foundations of machine learning and how to apply these to solve real-world data-driven problems. We will apply machine learning to numerical, textual, and image data. Topics will be drawn from perceptron algorithm, regression, gradient descent and stochastic gradient descent, support vector machines, kernels for support vector machines, recommendation systems, decision trees and random forests, maximum likelihood, estimation, logistic regression, neural networks and the back propagation algorithm, convolutional neural networks, recurrent neural networks, Bayesian analysis and naive Bayes, clustering, latent Dirichlet allocation (LDA), sentiment analysis, dimensionality reduction and principle component analysis, reinforcement learning. Prerequisite: For students following the 2019-20 or later bulletin, Introduction to Computer Programming, Calculus, (Probability and Statistics OR Theory of Probability). For students following the 2018-19 or earlier and are declared data science majors, Statistics for Business & Econ can be regarded as an alternative to Probability and Statistics OR Theory of Probability. Please contact your advisor for more information.
Reinforcement Learning (RL), a form of machine learning and a branch of Artificial Intelligence, enables an agent to learn in an interactive environment by trial and error using feedback from its own actions and experiences. RL seeks to learn a good policy for taking actions, using rewards and penalties as signals for positive and negative behavior. Modern RL problems are formulated as Markov decision processes with unknown environments.
There are two major sub-branches of reinforcement learning: tabular reinforcement learning for relatively small state spaces; and deep reinforcement learning, which combines deep learning and reinforcement learning, and is appropriate for environments with large (including continuous) state and action spaces. The course will cover both tabular and deep reinforcement learning. Probability theory and algorithms will be used throughout the course. Assignments will involve both mathematical derivations and programming assignments.
Natural language processing (NLP), a form of artificial intelligence (AI) that gives computers the ability to read, understand and interpret human languages, is one of the most important technologies that have made significant progress recently. NLP has been applied to many areas such as spoken dialogue system, machine translation, question and answering, machine reading, document summarization, and even music generation. Traditional NLP approaches involve rules that are handcrafted by linguists. On the other hand, modern NLP approaches are data-driven, trying to learn a model to minimize a target loss function over labeled or unlabeled training text. The course will cover various NLP techniques such as text classification, sequence classification, parse trees, and sequence-to-sequence generation from statistical or deep learning perspectives. Students will be expected to derive mathematical formulas, and code and tune NLP algorithms on datasets in homework assignments. Prerequisite: (1) Machine learning; (2) Probability and Statistics or Theory of Probability.
Computers are used to process signals, compose music, and perform with humans. Personal computers have replaced studios full of sound recording and processing equipment, completing a revolution that began with recording and electronics. In this course, students will learn the fundamentals of digital audio, basic sound synthesis algorithms, techniques for human-computer music interaction, and machine learning algorithms for media generation. In a final project, students will demonstrate their mastery of tools and techniques through a publicly performed music composition. Prerequisites:ICP OR ICS (best to have some experience in Music, or check with the instructor before enrolling).
This course offers an introduction to mathematical statistics. It covers the essential topics of statistics including point estimation, interval estimation, Bayesian inference, hypothesis testing, and linear and logistic regression. This class requires a good prior understanding of probability theory, calculus, and linear algebra. Prerequisite: Linear Algebra or Honors Linear Algebra 1, Multivariable Calculus or Honors Analysis II, and Probability and Statistics or Honors Theory of Probability.
This is an advanced topic course for undergraduate students interested in the modern mathematics of data science and machine learning. Tentative topics include dimension reduction and data visualization, the geometry of high dimensional data, and optimization-based data analysis. Topics may change every year to reflect the current research trends. The course requires an excellent understanding of advanced calculus, linear algebra, and probability theory. Programming skills and knowledge in optimization are strongly recommended but not required. Prerequisite: DATS-SHU 234 Mathematical of Statistics (used to be MATH-SHU 234).
Artificial Intelligence (AI) is reshaping business processes, creating disruptive innovations that change established industries and markets beyond recognition. The emergence of powerful algorithms, combined with recent growth in computational power and availability of massive amounts of data, enable companies to operate faster, make better decisions, automate processes, maximize revenue and customer engagement, among many other advantages. In this 7-week course we will briefly discuss some of the core principles underlying AI and then focus on a few selected applications of AI in business, such as predictive analytics for maximizing marketing and financial strategies, pattern recognition to understand customer behavior, and conversational AI and chatbots to improve engagement and customer experience. Last, AI also possesses significant limitations and poses new challenges with respect to fairness, biases, and automated errors. The course will conclude with a discussion of the main ethical issues and risks associated with AI technology. Prerequisite: Calculus and ICP. Student must have junior or senior standing.
The world we live in is built upon a myriad of networks: Human society is defined by our interpersonal relationships. Organizations are structured around interconnecting roles and lines of authority between workers, colleagues, and bosses. Global information is conveyed across a world-wide web of linked content. As we have witnessed recently, epidemics spread over a social network of contacts, in the same way in which we buy products as we are influenced by our peers. New sources of massive amounts of data fundamentally reflect interactions, and, in this context, networks are intuitive abstractions to model our social life, especially that mediated by technology. In networks, local interactions among members of small communities can often propagate and further affect the outcomes of an entire system. This course combines theories, models, and algorithms from computer science, economics, and the social sciences to analyze network data and find solutions to business problems. More information: https://shanghai.nyu.edu/is/course-spotlight-network-analytics Prerequisites: Introduction to Computer Programing (to manipulate network datasets), and Calculus.