20 Machine Learning Bootcamps and Courses Teaching the Art of Algorithms
Machine learning technology has the capacity to autonomously identify malignant tumors, pilot Teslas and subtitle videos in real time. Sort of. The term “autonomous” is tricky here, because machine learning still requires a lot of human ingenuity to get these jobs done.
It works like this: An algorithm scans a massive dataset. Engineers don’t tell it exactly what to look for in this initial dataset, which could consist of images, audio clips, emails and more. Instead, the algorithm conducts a freeform analysis. Then, based on that slice of information, it builds a model of how the world works, which it can use to, say, check pathology slides for early-stage breast cancer.
Even when it’s accurate enough to be packaged and sold, the algorithm continues to evolve — or “learn.” Whenever it makes a false assessment, it adjusts its underlying model accordingly. In that way, machine learning algorithms are groundbreakingly independent, capable of something many humans struggle with: self-improvement.
At the same time, the technology is a human creation. Humans build the algorithms and curate training datasets for them, which is no simple task. If an algorithm gets too much data, it can “overfit,” incorporating meaningless correlations into its model. With too little data, though, the algorithm works flawlessly on its training dataset only to flop in the real world. And once an algorithm has been trained and tested for accuracy, humans still have to engineer it into software, market it and — the list goes on.
Clearly, there’s plenty of work for people in this seemingly automated field, but landing a role like machine learning engineer requires cutting-edge technical knowledge. Hence the assortment of tech companies, bootcamps and universities that offer courses in machine learning and artificial intelligence. The programs vary widely in prerequisites, length and tuition — which means there’s something for everyone.
We rounded up 20 bootcamps and courses that teach the fundamentals of machine learning.
Springboard’s Machine Learning Bootcamp
What it is: This six-month bootcamp transforms software engineers into machine learning engineers. Starting with one or two years of coding experience, candidates learn the fundamentals of machine learning through a mix of digital materials and unlimited one-on-one mentorship. Students gain proficiency in the Python data science stack, study areas like natural language processing and, most importantly, practice production engineering — experience that is particularly valued by hiring managers. The final project, in fact, echoes the day-to-day of a machine learning engineer: To graduate, students must deploy a large-scale machine learning system.
Codesmith’s Software Engineering Immersive
What it is: Though it isn’t specifically a machine learning course, this 12-week bootcamp offered in Los Angeles and New York City takes students from zero to machine learning and beyond. (The only prerequisite: a high school diploma.) Starting with basic computer science principles, the curriculum progresses through front- and back-end development into a machine learning unit. There, students delve into Python and explore key data science concepts and libraries. Designed to prepare students for higher-level engineering roles, the course comes with lifelong job search support.
MIT’s Professional Certificate Program in Machine Learning & Artificial Intelligence
What it is: This program consists of a series of two- and three-day intensive courses, all taught by MIT professors on the university’s campus. Designed for data professionals with at least a bachelor’s degree, the interdisciplinary classes touch on math, statistics, computer science and programming. Graduation requires at least 16 total days of study, and two courses in particular: a two-day foundations course and a three-day advanced course, both focused on how machine learning can parse big datasets and text repositories. Students round out their schedules with electives on topics like computer vision.
Tech Elevator’s “Intro to Data Science”
What it is: In this workshop, which is offered in multiple, mostly-Midwestern locations and can be customized to specific workplaces and their needs, instructors walk IT professionals through the fundamentals of data science. The R-based curriculum, a collaboration between Tech Elevator and Pandata, covers data science topics that include the assumptions and concepts central to machine learning.
NYC Data Science Academy’s “Data Science with Python: Machine Learning”
What it is: During this 20-hour, part-time course — which meets near New York’s Penn Station on five Sunday afternoons — students learn to make predictions based on complex data sets. That means experimenting with discriminant analysis, support vector machines and other popular machine techniques under the watchful eye of professional data scientists who are drawn to the challenge of “big and messy” datasets. Admission requires familiarity with Python, the course’s lingua franca.
Online University Courses
Stanford’s “Machine Learning”
What it is: This Coursera course, taught by Coursera co-founder and Google alum Andrew Ng, starts simply enough — with a review of linear regressions, a.k.a. high school math. From there, though, the 56-hour curriculum delves into more esoteric topics, including cluster analysis and neural networking. Ng presents the course material in instructional videos, incorporating real-world case studies so students get a sense of how machine learning algorithms impact daily life. Students also complete supplemental readings and quizzes.
Columbia University’s “Machine Learning for Data Science and Analytics”
What it is: This edX course, which takes about 50 hours to complete, falls under the umbrella of ColumbiaX’s “Data Science for Executives” sequence. In that spirit, it’s less a deep-dive into the engineering process than an overview. The curriculum emphasizes machine learning applications in complex industries like healthcare, as well as typical workflows and techniques in the field.
Harvard’s “Data Science: Machine Learning”
What it is: In this edX course, students learn by doing — specifically, by building a movie recommendation system. Along the way, they learn about training data, popular algorithms and techniques for avoiding overfitting. A Harvard professor of biostatistics leads this introductory course.
University of Washington’s Machine Learning Specialization
What it is: This sequence of four Coursera courses begins with applications: what can this mysterious “machine learning” technology do? (Well, recommend products and value real estate, among other things.) Next, students delve into the mechanics behind these use cases. In the process, they learn to fit models that can classify data, retrieve relevant data and more. Each course blends video tutorials and quizzes.
California Institute of Technology’s “Learning from Data”
What it is: This 10-week course on Class Central covers the fundamentals of machine learning in 18 lectures, arranged in a narrative arc. First, the course establishes a definition of learning; then it delves into how that process can be automated. Individual lectures, available on YouTube, cover topics like the bias-variance tradeoff, Kernel methodology and more. Meanwhile, students find homework assignments and their final exam on the CIT course site — this digital course is almost identical to the on-campus incarnation.
Online Courses from Tech Companies
Google Cloud’s “Machine Learning on TensorFlow with Google Cloud”
What it is: In this sequence of five Coursera courses, students learn to develop machine learning models in Google Cloud — a platform with a hardware, tools and Tensorflow integration suitable for end-to-end engineering. The intermediate-level curriculum covers various Cloud capabilities, like model assessment and feature engineering, with a mix of videos, readings and hands-on labs. Students who plug away at this flexible sequence for 15 hours a week should complete it in about a month.
IBM’s “Machine Learning with Python: A Practical Introduction”
What it is: In this edX course, which takes 20 to 30 hours in total, students analyze real-life machine learning applications and experiment with models of their own. Led by Saeed Aghabozorgi, a data scientist at IBM, the course breaks down the differences between supervised and unsupervised machine learning, surveying techniques like the train-test split and assessing models by their root mean squared error.
Google’s Machine Learning Crash Course
What it is: This free course — or, as Google calls it, a “self-study guide”— consists of video lectures from Google researchers, case studies and more than 40 hands-on exercises. It takes 15 hours to complete, though it requires some familiarity with Python to start. The curriculum covers core machine learning concepts, training protocols and use cases, which are surprising and plentiful. For instance: Machine learning helped researchers analyze the political implications behind 18th-century writers’ metaphors.
Amazon Web Services’ “Machine Learning”
What it is: Before making more than 40 hours of instruction public, Amazon initially developed the curriculum for this free course to train its employees. Different “learning paths” prepare students for different technical roles, including data scientist and developer. In the end, participants can earn a certificate for their proficiency in crafting and deploying machine learning models in AWS.
Online Courses from Education and Tech Sites
Udemy’s “Machine Learning A to Z”
What it is: Comprised of more than 40 hours of video, readings and hands-on exercises, this introductory machine learning course covers a mix of theoretical concepts and practical applications. The goal is to make ideas like “convolutional neural networking” and “dimensionality reduction” ultimately feel like tools rather than gibberish. It’s also a rarity in that it requires no coding expertise. Instead, it comes with R and Python templates that students can modify and reuse in personal projects.
Sundog Education’s “Machine Learning, Data Science and Deep Learning with Python”
What it is: In this best-selling Udemy course, Frank Kane — who developed recommendation algorithms at Amazon and Imdb.com — teaches the fundamentals of machine learning in 13 hours of on-demand video lectures, interspersed with hands-on exercises. Students practice creating tools that automatically classify images, data and sentiments, for instance. The course also requires them to build bots and algorithms.
Udacity’s “Intro to Machine Learning”
What it is: This free course teaches students to look at data not just as information, but as fodder for algorithms. The curriculum mixes lectures from machine learning professionals with quizzes, guiding students through Python skills, approaches to dealing with data sets’ outliers and various types of algorithms — not to mention the art of picking the right one for a given project.
Fast.ai’s “Practical Deep Learning for Coders”
What it is: Designed for coders with a year or more of Python experience, this course requires two basic tools: NVIDIA GPU, an ultra-fast remote rental server; and Jupyter Notebook, an open source coding app. Equipped with these basics, students work independently through 14 free, project-based lessons. In an early one, participants learn to build image classification models that can distinguish real bears from teddy bears, among other functions.
Datacamp’s “Machine Learning Toolbox”
What it is: This four-hour crash course in machine learning — comprised of 24 videos and 88 interactive exercises — has been used as an employee training tool by major tech companies like PayPal and Dell. Led by two experts — one a top-ranked data scientist on Kaggle, the other a software engineer — the course consists of five R-based modules. The first, which is free, focuses on regression models; later modules focus on data preparation and model selection.
Kaggle’s “Intro to Machine Learning”
What it is: This “nanocourse” is the shortest on the list, clocking in at just three hours. Created by data scientist Dan Becker, who has worked with an array of Fortune 500 companies, the course gives a brisk overview of seven key topics in machine learning. They include the under- and over-fitting of models, random forest algorithms and more. Each lesson contains a text tutorial and a Python-based coding exercise. Once students complete the course, they can immediately put their newfound knowledge to practice on Kaggle’s platform, which hosts innumerable data science competitions.
Images via Shutterstock, social media and company websites.