Water's Home

Just another Life Style


1 Introduction to Deep Learning


Deep Learning has aready transformed the traditional internet businesses

What is a Neural Network

You probably find them to be most useful, most powerful, in supervised learning settings. Meaning that you’re trying to take an input x and map it some output y,

Supervised Learning with Neural Networks

It turns out that so far, almost all the economic value created by neural networks has been through one type of machine learning, called supervised learning. Possibly the single most lucrative application of deep learning today is online advertising.

  • Computer Vision
  • Speech Recognition
  • Autonomous Driving

It turns out that slightly different types of neural networks are useful for different applications. CNN (Convolutional Neural Network) RNN (Recurrent Neural Network)

  • Structured Data means basically databases of data.
  • In contrast, unstructured data refers to things like audio, raw audio, or images where you might want to recognize what’s in the image or text. Historically, it has been much harder for computers to make sense of unstructured data compared to structured data.

It turns out that a lot of short term economic value that neural networks are creating has also been on structured data, such as much better advertising systems, much better profit recommendations, and just a much better ability to process the giant databases that many companies have to make accurate predictions from them.

Why is Deep Learning taking off?

Go over some of the main drivers behind the rise of deep learning because I think this will help you better spot the best opportunities within your own organization to apply these to. More and more and more data have been collected so over the last 20 years for a lot of applications we just accumulate a lot more data more than traditional learning algorithms were able to effectively take advantage of. Hit very high level of performance :

  1. Train a big enough neural network
  2. Throw more data

If you don’t have a lot of training data is often up to your skill at hand engineering features that determines the performance. The performance depends much more on your skill at hand engineer features and other normal details of the algorithms and there’s only in this some big data regions very large training sets very large M regions in the right that we more consistently see largely neural nets dominating the other approaches and so

  • Data
  • Computation
  • Algorithmic Innovation

One of the huge breakthroughs in neural networks has been switching from a sigmoid function which looks like this to a ReLU function.

  • Ultimately the impact of this algorithmic innovation was it really help computation so there remains quite a lot of examples like this of where we change the algorithm because it allows that code to run much faster and this allows us to train bigger neural networks or to do so within reasonable amount of time even when we have a large network with a lot of data.
  • It turns out the process of training your network it is very intuitive often. you have an idea for a neural network architecture and so you implement your idea and code. Implementing your idea then lets you run an experiment which tells you how well your neural network does and then by looking at it you go back to change the details of your neural network and then you go around this circle over and over and when your neural network takes a long time to train it just takes a long time to go around this cycle and there’s a huge difference in your productivity building effective neural networks.

Faster Computation —> Iterate Much Faster —> Improve Ideas Much Faster

About this Course

Just use the multiple choice questions to check your understanding. And dont’ review, you can try again and again until you get them all right.

Course Resources

If you have any questions or you want to discuss anything with the classmates or … the best place to do that is the discussion forum.