Intro to Deep Learning (ML Tech Talks)

An overview of Deep Learning, including representation learning, families of neural networks and their applications, a first look inside a deep neural network, and many code examples and concepts from TensorFlow. This talk is part of a ML speaker series we recorded at home. You can find all the links from this video below. I hope this was helpful, and I'm looking forward to seeing you when we can get back to doing events in person. Thanks everyone!

Chapters:
0:00 - Intro and outline
3:58 - AI vs ML vs DL
7:55 - What’s representation learning?
8:40 - A cartoon neural network (more on this later)
9:20 - What features does a network see?
10:47 - The “deep” in “deep learning”
12:48 - Why tree-based models are still important
13:38 - How your workflow changes with DL
14:02 - A couple illustrative code examples
17:59 - What’s a hyperparameter?
19:44 - The skills that are important in ML
20:48 - An example of applied work in healthcare
21:58 - Families of neural networks + applications
28:55 - Encoder-decoders + more on representation learning
32:45 - Families of neural networks continued
35:50 - Are neural networks opaque?
38:29 - Building up from a neuron to a neural network
49:11 - A demo of representation learning in TF Playground
53:24 - Importance of activation functions
54:36 - What’s a neural network library?
58:43 - Overfitting and underfitting
1:02:38 - Autoencoders (and anomaly detection) screencast and demo
1:12:13 - Book recommendations

Here are three helpful classes you can check out to learn more:

And here are all the links to demos and code from the video, in the order they appeared:

[TensorFlow]

Here are three helpful classes you can check out to learn more:


And here are all the links to demos and code from the video, in the order they appeared:

[JBGordon]

It was fun recording this! I hope it's helpful to you. I know there are many intro to dl talks :) A good strategy you can use to learn a topic is to leverage talks (and books!) by different people on the same idea in parallel. Everyone covers it a bit differently. Some of their explanations will click for you, and you can merge them into your own understanding. I left links to a bunch of my favorite courses + books in the video description for you (they're really great!), so you can dive deeper.

[farhanfuadabir]

It is by far the best introduction to Deep Learning I had. Thanks.

[xedski]

The best I've seen to date. Not about coding, not about matrix algebra, it's about concepts behind what is going on with tons of parameters whil eventually result in a model.

[donquixoterosinante]

I love his explaination and genuine kindness he show throughout the video

[yk]

The sigmoid function (in the context of logistic regression) is not just interpreted as probability, it truly yields the probability, though the fact that it can be analytically derived is mostly overlooked in ML courses.

[ayoolafakoya]

Fantastic Intermediate introduction to Deep Learning .

[masudRana-brev]

A complete presentation indeed. The way you followed was simple and easy for everyone. Thank you very much for sharing your knowledge .... and you always had a smiling face that made the presentation more attractive to me. Again thanks a lot

[soukisama]

Thank you! High level content explained in a simple and direct way. Bests!

[Trident]

That was a lot of ground covered in in 75 minutes! Bravo Sir, your skills as a teacher are being honed in front of our eyes!

[prakashsuthar]

Thank you so much for recording excellent presentation. You have covered all key concepts exceptionally well. Please continuing sharing your knowledge.

[nasilasy]

Was going to just watch the first few minutes and come back to it later but end up watching the whole video. The best high level introduction to deep learning and it's accompaning concepts. Thanks for sharing this valuable resource.

[borin]

This is the best presentation that mostly covers all the things in Machine Learning. Thank!

[notsure]

I have never seen the IT community so excited. Thank you all around the world..😁

[uty]

this is unbelievably good, makes neural network interesting, even to an old man like me haha

[mohammedqasim]

Very well explained, thank you for doing this and best of all the last few mins on the reference books! super like!

[kunalsoni]

great explanation :) I'm very happy to watch this

[ekaterinadranitsyna]

Josh, thank you very much for this overview of neural nets! It's really useful to have it all in one place!
I have a tangent question: is it possible to run TensorFlow on AMD GPUs on Windows? Would it help to install Linux subsystem and ROCm or it wouldn't work?

[SuperNelmer]

We miss you at our GDG meetups Josh! ToT

[halilibrahimaysel]

Thank you so much for this amazing recording! I was just wondering if you can recommend any paper for interpretation with "Integrated Gradients". It will help me a lot!