Deep Learning at Udacity Evolves

Deep Learning at Udacity Evolves

Enrollment is now open for the latest Deep Learning Nanodegree program! Discover amazing new content, and explore your future in Deep Learning, today!

Udacity - Deep Learning Nanodegree program

The Deep Learning Nanodegree program was one of the first Udacity programs built as a direct and immediate response to the very latest advancements in the field of AI, and as such, it was an early and unprecedented opportunity for aspiring learners to master valuable and in-demand deep learning skills. Deep learning is such a dynamic and rapidly-advancing field, and it has been a delight to see so many students learn and grow with this field. Thousands of students have graduated from the program, and many have gone on to great careers at companies like OpenAI, NASA, and more—not to mention the amazing personal projects our graduates continue to build!

As researchers learn more about deep learning, and as the technology evolves, our curriculum must advance as well. The rapid pace of change in this field means that we’re constantly updating and enhancing the content in this program, in order to consistently ensure our students always have the best educational experience possible. Staying up-to-date with a field this innovative isn’t easy, but our commitment to doing so is a big part of why this program is so special.

In this post, I’m going to share some exciting new updates to our Deep Learning Nanodegree program: the use of an additional deep learning framework: PyTorch, a new section on Model Deployment, and a new lesson on Image-to-Image transformation!

Udacity - Deep Learning - Image-to-Image Transformation

Deep Learning with PyTorch

There are a number of frameworks available to help you construct and train deep learning models. The most popular are TensorFlow, Keras (which has been wrapped into TensorFlow), and PyTorch. In this update of the Deep Learning Nanodegree program, we’ve built our content with PyTorch as well as TensorFlow. We’ve found that PyTorch is better pedagogically, in that it maps closer to deep learning concepts while also having a shallower learning curve for those experienced with Python. PyTorch lets you play to the strengths of Python: readable code, speedy development, and flexibility. On the other hand, Tensorflow has been heavily optimized for running in production environments and in distributed environments with dozens or hundreds of CPUs and GPUs. As such, our main examples are in PyTorch but we’re also providing the same content in TensorFlow.

Model Deployment

Udacity’s Nanodegree programs are designed for you to learn valuable skills that are used in industry, today. As more and more companies look to build AI products, there is a growing demand for engineers who are able to deploy machine learning models to global audiences. To help prepare you to take advantage of this demand, and qualify for roles of this kind, we are adding a new section on model deployment and model serving to the Deep Learning Nanodegree program. Here, you’ll get hands-on experience deploying and monitoring a model using PyTorch and Amazon SageMaker. By teaching these essential skills, we are preparing our students to be indispensable members of AI product teams.

Unpaired Image-to-Image Translation

[image: Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks, J. Zhu, T. Park, P. Isola, A. Efros, 2018]

Image-to-Image Translation

This course will cover the latest in deep learning architectures used in industry, including architectures called Pix2Pix and CycleGAN. These models approach the challenge of image-to-image translation tasks, such as transforming images from winter to summer or turning sketches into realistic images.

We also focus on Generative Adversarial Networks (GANs). GANs are a relatively new invention, introduced by Ian Goodfellow in 2014, and Udacity has partnered with Ian to provide instruction on this unique class of artificial intelligence algorithms.

The opportunity to partner with experts in both industry and academia is an important benefit for our students, as it enables us to provide you with the most in-depth looks at the latest technologies. Ian’s 2014 GAN paper spurred on even more GAN research, and we’re excited to have another expert on board to enhance your learning experience. Jun-Yan Zhu is a researcher at MIT’s CSAIL, and he’ll be teaching you about his recently-published work on CycleGANs.

The Power of Deep Learning

Deep learning is constantly evolving, and this field has shown continuous growth over the past few years. There has never been a better time to start learning about the deep learning models that are changing the way we work. It is really up to us as learners and teachers to shape how this technology advances, and we can do so by learning about the latest deep learning techniques and coding our own models. If you are curious about the subject, and are interested in applying deep learning skills to personal or professional projects, then this course is for you!

Enroll in the Deep Learning Engineer Nanodegree Program today, and experience the power of deep learning!


Please enter your comment!
Please enter your name here