Link to Content:

Deep Learning Nanodegree

Created/Published/Taught by:

Udacity

Siraj Raval

Mat Leonard

Brok Bucholtz

Content Found Via:

Udacity

Free? No

Cost: $0.00

Tags: artificial intelligence / deep learning / machine learning / supervised learning

Difficulty Rating:

From Udacity:

“Build your Deep Learning Foundations, and earn your Udacity credential!

Artificial Intelligence is transforming our world in dramatic and beneficial ways, and Deep Learning is powering the progress. Together with Siraj Raval, Udacity provides a dynamic introduction to this amazing field, using weekly videos, exclusive projects, and expert feedback and review to teach you the foundations of this future-shaping technology.”

Siraj’s bio: “I’m a Data Scientist, bestselling author, and YouTube star. I make videos that teach people how to use machine learning to create game bots, chatbots, self-driving cars, programs that create art and music, stock prediction models, and much more. I’m proud to be an exclusive Udacity partner, and excited to be your host for this amazing program.”

Syllabus by week:

- Introduction to Deep Learning
- Model Evaluation and Validation
- Graph computations
- Intro to TensorFlow
- Deep Neural Networks
- Convolutional Networks
- Recurrent Neural Networks
- Word Embeddings
- Using TensorBand
- Text Generation
- Sequence to Sequence Generation
- Transfer Learning
- Reinforcement Learning
- Autoencoders
- Generative Adversarial Networks (GAN)
- Image Generation
- One-shot Learning

Recommended Prerequisites: This program has been created specifically for students who are interested in machine learning, AI, and/or deep learning. Students who are interested in enrolling must have intermediate Python programming knowledge, experience with Numpy, experience using Anaconda and Jupyter Notebooks. Additionally, students must have the necessary math knowledge including: algebra and some calculus – specifically partial derivatives, and matrix multiplication (linear algebra) to be successful.

Go to Content: Udacity Presents Siraj Raval’s Deep Learning Nanodegree Foundation Program

## By pmarelas June 4, 2017 - 8:21 pm

Last week I completed the Udacity Deep Learning Foundations Nanodegree. As the name implies this course teaches you the foundations of Deep Learning. The course is made up of a mix of theory, humour and projects. The theory is delivered by Udacity employees and external tutors, while the humour is delivered by Siraj Raval. Some people may find Siraj’s style difficult to stomach but personally I found his tutorials entertaining and enlightening. Siraj helps connect the subject matter to solving real work problems, which I’ve found lacking in other courses.

So, what did I learn? I learnt the algorithms and methods that underpin Deep Learning, including feed forward networks, stochastic gradient descent, back propagation and the role of activation functions. The course gives you a good dose of math required to supports these methods.

With Deep Neural Networks, the network architecture is equivalent to feature engineering in traditional Machine Learning. You will learn about different network architectures and where to use them including Neural Networks, Convolutional Neural Networks, Recurrent Neural Networks, Reinforcement Learning, Autoencoders and the more recent Generative Adversarial Networks.

To complete the course, you must submit five projects. Each project is accommodated by one or more tutorials. The projects are written in python using jupyter notebooks and a combination of numpy and tensorflow. The goal of the projects is to ensure you understand and can apply each subject matter.

In my experience the first week was the toughest. You learn to implement the fundamental mathematical concepts that underpin Deep Learning in python, with no help from machine learning libraries. Luis Serrano teaches you the perceptron and Mat Leonard teach you how to implement your first neural network using python and numpy.

The question you probably want to know is would I recommend this? My answer is twofold. If you have no background in Machine Learning I would advise you to study a traditional Machine Learning course before jump straight into Deep Learning. This will help you fill in some of the gaps that may arise. For myself, I reviewed several materials before jumping into Deep Learning. They were:

• Udacity’s Intro to Machine Learning contains a mix of theory and practical projects

• machinelearningmastery.com by Jason Brownlee has some excellent self-paced books that are reasonably priced. Also contains a mix of theory and practical implementations, even in Excel!

• Machine Learning by Stanford is more theory than practice available on ITunes

Once you have the traditional Machine Learning concepts covered then I highly recommend the Udacity Deep Learning Foundations Nanodegree.

I’m giving the course 4 of 5 stars because the content was being produced as it was being consumed by students. This was circumstantial as I was part of the first class of students to take the course. You should not experience the same issue.

Advanced Beginner- familiar with data science / new to this topic / some prerequisitesIntermediate - had relevant prerequisites / was familiar with this topic

(4) A Lot

Detailed Content Survey:Needed to gain or strengthen general data science knowledge

This is one of my primary go-to content sources

Yes

All of it

Understanding of traditional Machine Learning concepts.

Python programming.

Python, GPU (can use cloud)

* Udacity’s Intro to Machine Learning contains a mix of theory and practical projects

* machinelearningmastery.com by Jason Brownlee has some excellent self-paced books that are reasonably priced. Also contains a mix of theory and practical implementations, even in Excel!

* Machine Learning by Stanford is more theory than practice available on ITunes

Understanding the different neural network architectures and how to apply them.

Practical experience implementing deep learning with tensorflow.

Projects

Some content was delayed

Don't be discouraged in the first week. Draw on the other content described under recommendations for other content. Also, see help from the Udacity forums and slack channels they setup for this course.

I followed a sequential approach avoiding the temptation to skip lessons and move onto other projects.

The course runs for 6 months. I completed within first 4 months and spent about 8 hours a week. Expect to spend more time to get through the first few lessons.

Advanced Beginner