Deep learning is machine learning with many layers in the model. Instead of you telling it which features to look at, the network works them out itself from the data.
That is why it works so well on images, speech and text: nobody could write down the features for those by hand.
Deep Learning is all exciting! Deep Learning can be used for making predictions, which you may be familiar with from other Machine Learning algorithms.
Becoming good at Deep Learning opens up new opportunities and gives you a big competitive advantage. You can do way more than just classifying data..
Deep Learning Applications
Deep Learning can be applied in many industries: Consumer, Industry, Art, Finance, Science, Robotics, Energy, Transportation and more.
Applications of Deep Learning in the real world are:
| Task | Details |
| Recognizing faces | Recognize faces in images or videos |
| Object Recognition | Neural Networks can recognize objects in images, in fact, better than humans. |
| Caption Generation | Given input of an image, it can output a text describing whats happening in the image |
| Deep Speech | All of the big companies have a Speech Recognition system that is based on Deep Learning. Given a sound file or real time sound, the software can convert it to text. |
| Language Translation | Given an input language, say English, a neural network can translate it to another language in a natural way. Not just that: think real-time translation. |
| Data Centers | Optimize cooling of data centers. Google used it to save billions of dollars based on an algorithm! |
| Medical: Automated Detection | Automated Detection of Retinal Disease or other diseases |
| Self driving cars | Complete autonomous driving |
More on Deep Learning with Tensorflow.
Why depth matters
- The first layer picks up simple things, edges and colours.
- The next layer combines those into shapes, corners and textures.
- The layers after that put the shapes together into parts of an object.
- The last layer gives you the answer.
You do not set this up. The network learns the filters while it trains.
Because the features come out of the training data, a deep model needs a lot of data to work well. A few thousand images is not much. It also needs a GPU to train in a reasonable time, which is why most people start with a smaller model.
For a lot of problems a classic model is still the better choice. A decision tree or logistic regression on a table of numbers will train in a second and often gives a better score. Reach for deep learning when you have images, audio or text and a lot of them.
Reading helps, writing fixes it. PyChallenge has exercises on this and you can try them right now.