Data Science (DS-GA)

This required course for the MS in Data Science should be taken in the first year of study. It introduces students to basic algorithms and software tools, and teaches them how to deal with data, representing data, and methodology. Hands-on experience using Torch, a software system being developed at NYU and other research centers that has a large data science library, will also be provided. Prerequisites include: Basic Probability or Statistics (undergraduate level), Calculus I, Linear Algebra, some experience in programming: Java, C, C++, Python, R, Lua, Ruby, OCaml or similar languages (equivalent to two introductory courses in programming, such as “Introduction to Programming” and “Data Structures and Algorithms”). Some prerequisites may be waived with permission from the instructor.

This required course for the MS in Data Science should be taken in the first year of study. It covers fundamental concepts in probability and statistics from a data-science perspective. Prerequisites include: calculus I, linear algebra, basic programming skills and some experience with probability or statistics at an undergraduate level. Some prerequisites may be waived with permission from the instructor.

This required course for the MS in Data Science should be taken in the first year of study. It covers a wide variety of topics in machine learning, pattern recognition, statistical modeling, and neural computation. It covers the mathematical methods and theoretical aspects, but primarily focuses on algorithmic and practical issues.

This required course for the MS in Data Science should be taken in the first year of study. It covers methods and tools for automatic knowledge extraction from very large datasets. Methods include on-line learning, feature hashing, class embedding, distributed databases, map-reduce framework, CUDA GPU programming, and applications.

This course covers graphical models, causal inference, and advanced topics in statistical machine learning.

This required course for the MS in Data Science should be taken in the second year of study. The purpose of the capstone project is to make the theoretical knowledge acquired by the students operational in realistic settings. During the project, students see through the entire process of solving a real-world problem: from collecting and processing real-world data, to designing the best method to solve the problem, and implementing a solution. The problems and datasets come from real-world settings identical to what the student would encounter in industry, government, or academic research.

The class will teach students about programming for applications in data science. Students will study the Python language and packages including tools for array operations, table manipulations, visualization, and data extraction. Through a focus on examples, students will learn about query languages, version control systems, and web frameworks. Experience with debugging, testing and documenting programs will enable students to code in integrated development environments and command line interfaces.

This course concerns the latest techniques in deep learning and representation learning, focusing on supervised and unsupervised deep learning, embedding methods, metric learning, convolutional net and recurrent nets, with applications to computer vision, natural language understanding, and speech recognition. The pre-requisites include DS-GA 1001 Introduction to Data Science and DS-GA 1003 Machine Learning

This course provides data science students with an opportunity to apply the knowledge gained in the course work to one or more practical problems in industry, medicine, government, or research. Students can only take this course at most twice.

This independent study course provides students with the opportunity to work one-on-one with a faculty member on a particular topic or project. The learning objective of the course is to build or to strengthen data science skills through focus on specific issues of interest to the student.

How should human languages be understood and analyzed? This course examines modern computational approaches based on representation learning for understanding, processing and using human language. These include neural network-based deep learning methods and vector-space models of word meaning, and together will give the tools to build state-of-the-art models for hard language understanding tasks like translation. Prerequisites include: undergraduate level probability and statistics, undergraduate level linear algebra, undergraduate level calculus, DS-GA 1003 or CSCI-UA.0480-007

Since at least the proposal of the Turing test, building computational systems that can communicate with humans using natural language has been a central goal for Al research. Understanding real, naturally occurring human language is the key to reaching this goal. This course surveys recent successes in language understanding and prepares students to do original research in this area, culminating with a substantial final project. The course will focus on text, but within that will touch on the full range of applicable techniques for language understanding, including formal logics, statistical methods, distributional methods, and deep learning, and will bring in ideas from formal linguistics where they can be readily used in practice. We'll discuss tasks like sentiment analysis, word similarity, and question answering, as well as higher level issues like how to effectively represent language meaning.

This course provides a rigorous introduction to mathematical tools for data science drawn from linear algebra, harmonic analysis, probability theory, and convex analysis. The main topics are the singular-value decomposition (SVD), the Fourier series, randomized projections, the randomized SVD, convex optimization, duality theory and nonconvex optimization. The material is motivated by multiple data-analysis applications including dimensionality reduction, collaborative filtering, sound and image processing, magnetic-resonance imaging, sparse regression, compressed sensing, and topic modeling.Prerequisites include: probability, calculus, linear algebra, experience in programming

This course covers the basics of optimization and computational linear algebra used in Data Science. About 66% of the lectures will be about linear algebra and ~33% about convex optimization. The first 5 lectures will cover basic linear algebra: vector spaces, linear transformations, rank, norms and inner products, eigenvalues and eigenvectors. Then we will study applications: Markov chains and PageRank, PCA and dimensionality reduction, spectral clustering, linear regression. Lastly, we will go over convex functions, optimality conditions for constrained optimization and gradient descent.

Course introduces students to quantitative texts-as-data analysis from an applied perspective. Course covers, inter alia, metrics of association between texts, burstiness of words and concepts, measurement of complexity and readability, scaling of political texts, automatic event extraction, dictionary methods for estimating sentiment, clustering, Latent Semantic Analysis, machine learning applications, topic models and LDA. Course also includes special topics such as the estimation of personal characteristics from writings, 'stylometrics' and detection of false statements. Course assumes no prior knowledge of texts-as-data work, though requires proficiency with the R statistical language and programming environment along with an understanding of elementary statistical theory and regression analysis.

This course provides a survey of computational approaches to understanding human intelligence and cognition. Both psychologists and data scientists are working with increasingly large quantities of human behavioral data. Computational cognitive modeling is the project of understanding behavioral data (and the mind and brain, more generally) by building computational models of the cognitive processes that produce the data. The course will cover the goals, philosophy, and technical concepts behind computational cognitive modeling, including model fitting and evaluation. Ideally, students will leave the course with a richer understanding of how computational modeling advances cognitive science, how cognitive science can inform research in machine learning and artificial intelligence, and how to fit and evaluate cognitive models for understanding behavioral data.

The first wave of data science focused on accuracy and efficiency – on what we can do with data. The second wave focuses on responsibility – on what we should and shouldn’t do. Irresponsible use of data science can cause harm on an unprecedented scale. Algorithmic changes in search engines can sway elections and incite violence; irreproducible results can influence global economic policy; models based on biased data can legitimize and amplify racist policies in the criminal justice system; algorithmic hiring practices can silently and scalably violate equal opportunity laws, exposing companies to lawsuits and reinforcing the feedback loops that lead to lack of diversity. Therefore, as we develop and deploy data science methods, we are compelled to think about the effects these methods have on individuals, population groups, and on society at large. Responsible Data Science is a technical course that tackles the issues of ethics, legal compliance, data quality, algorithmic fairness and diversity, transparency of data and algorithms, privacy, and data protection.

This course presents fundamental tools for characterizing data with statistical dependencies over time, and using this knowledge for predicting future outcomes. These methods have broad applications from econometrics to neuroscience. The course emphasizes generative models for time series, and inference and learning in such models. We will cover a range of approaches including Kalman Filter, HMMs, ARMA, Gaussian Processes, and their application to several kinds of data.

In this course, we will examine a range of advanced techniques for improving the performance of Python programs, including the use of parallel computation and GPU acceleration. We will also investigate how Python can be used for big data analysis using frameworks such as Apache Hadoop and Apache Spark. Students will have the opportunity to employ these techniques and gain hands-on experience developing advanced Python applications. The course will take a student-centered, active learning, approach to teaching this material. Class will typically consist of a short introduction to programming techniques, followed by hands-on computing exercises.

This course provides rigorous tools for the mathematical analysis of statistical procedures in data science. Topics include hypothesis testing, confidence sets, regression, matrix estimation, and classification. The course will focus both on classical asymptotic theory and on modern non-asymptotic techniques and theorems suitable for data science applications.

Reviews a number of important algorithms, with emphasis on correctness and efficiency. The topics covered include solution of recurrence equations, sorting algorithms, selection, binary search trees and balanced-tree strategies, tree traversal, partitioning, graphs, spanning trees, shortest paths, connectivity, depth-first and breadth-first search, dynamic programming, and divide-and-conquer techniques. Prerequisites: At least one year of experience with a high-level language such as Pascal, C, C++, or Java; and familiarity with recursive programming methods and with data structures (arrays, pointers, stacks, queues, linked lists, binary trees).

The research rotation course gives PhD students exposure to the research being conducted by CDS faculty. The objective of this course is to broaden students’ perspective and make them better rounded data science researchers. During this semester-long course, students will design and carry out original research in a collaborative setting with faculty who will monitor progress on a weekly basis and assign a pass/fail grade at the end of the semester and submit a brief report to the DGS.

This course is a 7-week course for CDS students, particularly PhD students, consisting of two separate and discrete Short Course components -- a 4-week Academic Writing course and a 3-week Great Presentations course. The Academic Writing course component is an intensive introduction to the principles of excellent rhetorical writing with a focus on the development of a clear, interesting, and rigorous science text, the construction of logical arguments, and the reporting of data, as well as the important concepts including reader-oriented writing, genre, precision, tone, the composing process, and strategies useful for redrafting and editing. Some of the sub-genres we analyze and practice include introductions, data commentaries, results/discussion, conclusions, and abstracts. We also practice other professional texts including requests for funding and professional email texts. The fundamental principles discussed and practiced in the Great Presentations course component can be applied in a variety of contexts including the short research talk, a lab talk, a formal conference presentation, poster presentation, job talk, interview, industry pitch. We talk about how to construct a logical and interesting presentation story, the design and best use of visuals, transitioning through the story and visuals, fluent delivery, connecting with the audience, timing, coordination of movement with content, and key linguistic elements such as volume, pitch range, intonation, and ends of utterances. We also look at the art of asking and responding to questions. Students will give talks of varying lengths followed by detailed feedback from the instructor.

Data Science is a new discipline. This brings challenges and opportunities. On the one hand, its boundaries are poorly defined; on the other hand, this very fact provides unusual latitude for us to establish its intellectual core in real time. This course is about that core. That is, we aim to provide students with an overview of Data Science as an endeavor: its origin, scope, techniques, debate and future. Consequently, this course is both broad and deep. It is broad in that it covers considerable ground in statistics and computer science in a short space of time. The central goal therein is to provide a basic and common vocabulary for students coming from multiple disciplines, enabling them to understand work in the field and to communicate their own work. The course will provide practical hands-on experience with Python and its associated data analysis libraries for this purpose. The course is deep in that we will cover some fundamental Data Science ideas in some detail—both technically and philosophically. This extends to the ethics of Data Science work.

This course is always offered in sections, each section around a special topic. The special topics will vary from time to time depending on the availability of suitable instructors. The format will vary by the topic, but will usually include an introduction to the topic and an overview of advanced research in the topic.