Deep learning nanodegree foundation | Hacker News Reader

Deep learning nanodegree foundation

udacity.com

83 pointszuzoovn34 comments

Hi all,

My name is Dhruv and I work on the AI programs here at Udacity. If you have any questions, feel free to ask them here or email me at dhruv@udacity.com. I'll do my best to answer everything! You can see our curriculum in depth here: https://medium.com/udacity/deep-learning-nanodegree-foundati...

In the same spirit: my name is Oliver and I lead the Udacity open source self-driving car project! AMA, or email me at oliver@udacity.com. I love all things AVs and deep learning!

This is the only thing that makes me said: "At this time we are not awarding scholarships for this program."

There seems to be a pretty wide spectrum of abstraction for DL tools, from writing by hand to Theano to Keras, for example. Will this course focus on any tool in particular?

Hey @wcbeard10,

This course will focus primarily on TensorFlow. We're doing so because, at the moment, it's the most popular Deep Learning framework and provides you enough flexibility to explore some of the newer network architectures we focus on towards the end of the program.

~80% focusses on TensorFlow, but we are working on Keras content in other Nanodegrees!

Nice to see you here on HN, Oliver :)

Would you say that this program is subset of material covered in car nano-degree program?

This is all new material!

Hi @halite,

I'd say that covers a larger breadth of Deep Learning than the Self-Driving Car Nanodegree program. For instance, this program will cover Recurrent Neural Networks, Reinforcement Learning, Autoencoding, and other Deep Learning applications not covered in the Self-Driving Car Nanodegree. In the Self-Driving Car Nanodegree, we primarily focus on Deep Learning applications in Computer Vision. Here we'll cover applications in speech, computer vision, game-playing and other areas.

Thanks this is useful info. I'm currently selected in that program and was wondering how's this different from that. I'm probably going to do the self-driving program this year.

Is it mandatory that I have to finish the course with 17 weeks or whatever the time duration is? Let's say if it takes me year to finish the course is this acceptable? I am interested to learn Deep learning and gain knowledge to concepts and new materials which Udacity is providing. I'm going to be most likely a passive learner not in a pursuit of degree.

Also will there is new materials added periodically or is this for a short duration of time frame?

Hi Akshay,

You have 6 months in total to finish all the content. Siraj is also posting some segment of the content for free on YouTube for everyone to see. The new material is added weekly!

hi! For some of us who have never taken a udacity course can you maybe give a bit of background information about what the weekly time expectations are? And also maybe a quick outline of the format for interaction? (or a link?)

Hey @dnautics! The main deadlines are around 5 projects that you will do over the course of the 17 weeks. We have suggested deadlines for them but the only requirement is to pass all of them within the 6 month date of the program. There will be other weekly assignments but those are meant for you to go at your own pace. We've designed this so that those who have a full time job and family can stay on track while making progress.

1. How much power would we need? I'd assume a Thinkpad X230 won't do. If we need a moderately powerful system, do you provide AWS (or equivalent)?

2. I watch Siraj's videos. Although they are fun and useful, they are very short and resemble a recipe in a cookbook (no offense intended). Is the course going to be the same as videos? Or is it going to be discussing all the necessary mathematics/statistics?

3. I couldn't find detailed information about the instructors' background. Are they able answer questions from the deep learning book or Sutton&Barto, if one were to read them alongside the course?

Hi @aalireza,

1. We'll provide instructions to use AWS (65 cents an hour) and are working with AWS to get students credits. We've done this in the past so there's a decent chance we get this but no guarantees yet.

2. The course is going to be Siraj's videos supplemented with additional content to go deeper into the math/statistics and to actually do problem sets and projects. So Siraj is meant to be the starting point from where we go deeper.

3. Mat Leonard will be the main instructor and did is PhD in Physics and Statistical Neuroscience at Berkeley. He's done his fair share of reading from books in the field and will be able to answer your technical questions.

Hi @dhruv,

So we require AWS for sure to complete this course right? Can you suggest the system configuration needed for this course?

how much time is one expected to spend on this course each week?

Hey @qrco, you'll be expected to spend 4 hours a week.

What does "Live session" mean? I can't find anywhere on Udacity where this is described. It seems like it is a selling point of the program, but why it is a selling point, seems unclear on the on boarding page.

Hi @wodenokoto - sorry for the lack of clarity. The Live Sessions will be a live Youtube stream held by Siraj where he answers your questions as you ask them in the session.

Will any of the official content (eg. lectures) for the nanodegree be offered free to those not enrolled?

Hi Dhruv, thanks for being available to answer some questions. Do you think that someone from a design background with some JS experience is able to do this course?

Siraj's videos are great! I'm excited to see him teaming up with my favorite Education company and creating a new ND!

This program looks promising. I have a few questions:

1)It is mentioned that it takes about 3-4 hours per week. Does this mean that the course covers not many concepts. Also are the projects like- this code is given. Add a few lines to make it run?

2)Can I submit files directly using my own laptop, or is using AWS compulsory?

3)Are these projects too - Make a network that generates songs and poems?

4)Can I get my resume reviewed by the experts like in the other nanodegrees?

It would be great if you could answer these questions. I am definitely thinking of joining this amazing course.

Can you provide details on what type of computing environment will be used in the course? Will it be based on something like Jupyter notebooks, GPUs on AWS, for example?

And will be there be some type of discussion forum, like the one in the other popular deep learning course fast.ai?

Hey @pchal,

We'll mainly use Python with Jupyter notebooks and will gear the program towards GPUs on AWS. We'll focus on using Tensorflow as our main tool. There will be a discourse forum as well as a slack channel for students to interact and help eachother.

Insta-buy! Currently doing Hinton's Neural Networks course on Coursera. Audited it when it was first offered, but am doing the assignments, for real, this time.

What's good @dhruvp, I see that we have 6 months in total to complete the course. Does that start from the time of sign up or when we actually start the course? I ask because I'd like to take advantage of the great price but I know my workload won't permit me starting for at least a month.

What's happening @king_kerr (Go Dubs). It starts from the first day of class which is Jan 20th.

Thanks!

How much of the course is ready to go? If a person could dedicate 20+ hours a week is the material all available or would you have to wait for the material to roll out? (i.e. is this binge-ready?)

The "nanodegree."

If anybody ever tried to use that as a credential, they'd waste more words trying to explain what it is, and what it's worth, than if they just told you what they actually knew verbatim.

I have a nanodegree in microwaving hot pockets. Ask me to defend its intrinsic value.

Reply on news.ycombinator.com