 Deep learning in healthcare using the most popular neural network architecture, TensorFlow, brought to you by the School for Data Science and Computational Thinking right here at Stellenbosch University. My name is Dr Jean Klopper and I'm a research fellow at the School for Data Science and Computational Thinking. I'm the creator and instructor for many online courses, the largest of which has more than 100,000 participants from all over the world. I've also authored a textbook on statistics. One thing is for sure, I am passionate about learning from data and about showing others the beauty of extracting knowledge from data. I hope that this passion shines through during this course. I really want you to be passionate about it too. Now, week is a short time in which to learn any subject. In deciding what to put into this course, I want it to be cognizant of the audience. We are all domain experts in our fields or at least learning to become experts and for most of us there is just so much data from which we can extract knowledge. I want to leave you at the end of this week with a full understanding of the power of deep neural networks. This means that this is not an easy course. Learning any new topic takes time and effort and practice. We create deep neural networks using computer languages and writing code. A computer language is much like a spoken language. You don't just pick up a new language and become fluent in it in a week. This course aims to be both a reference work packed with information and also a personal guided journey. I have created video lectures, multiple PDF documents, practice exercises, solution sets and we're going to have live online sessions. I really want to support you during the start of your journey. There is no way that I want to leave you with some superficial useless start. So I need you to stick with it. Remember there is no pressure during this week. No expectation to understand everything. We are not aiming to develop the ability to have a deep conversation on the meaning of life in Latin at the end of this week. The aim is rather for you to have a full understanding of the potential of deep neural networks. I want to guide you towards a future where you can join this massive community of domain experts employing the power of deep learning in their fields. Python has become the leading language in machine learning and artificial intelligence. It is an easy to learn language with a very clear syntax. Once you know some Python code, you can pretty much guess what a new piece of code should look like. Answers to Python related questions are also everywhere. A quick search of Google will show pages and pages of links to tutorials, videos, discussion boards, and so many other resources that will answer all your questions. You will pretty much never get stuck when you need an answer about Python or TensorFlow for that matter. The computer language needs a program into which you type your code. In this course, we are going to make use of Google's Collaboratory. The co-lab for short is very similar to a Google Doc. Once you have a Gmail account, you have access to Google Docs. A co-lab notebook looks like a Google Doc. It is just a blank web page into which you can write normal words and sentences, format titles and subtitles and add pictures and videos. You can also enter Python code though and see the results of your code. Using Google co-lab means you don't have to install Python or TensorFlow on your local machine. Now, it is completely possible for you to do so if you want to though. If you have questions about this, we will discuss it during the live sessions. I mentioned the word notebook. This is what a Google co-lab file is called. There is a series of notebooks that you will have to upload to your Google Drive. Later in this video, I will show you how to do this. You will also have access to a set of PDF documents that can serve as reference material to read if you are not into watching lectures. I will also make use of a second online service called Kaggle. Much like Google co-lab, it is free to open an account and Kaggle also generates Python notebooks. We will use a Kaggle notebook to design and train a deep neural network to distinguish between benign and malignant skin lesions using thousands of pictures of skin lesions. The course is structured so that you need to watch the required video lectures and or read the documentation each morning or at least if you want to the night before. There is a set of exercise notebooks that you can attempt after each lecture and at a specific time every afternoon, we will have a live online session where we work through the exercises and discuss related topics. Deep neural networks are a form of machine learning which is in turn a type of artificial intelligence. It has gained massive popularity and is at the heart of self-driving cars and recommended systems that suggest which TV show you should watch next or what other item you might be interested in when online shopping. Now there are 14 chapters in this course and I want to tell you a little bit about each of them. Python is zero indexed. In other words it starts counting at zero so in chapter zero I introduce Python and Google Colab. We start with some simple arithmetic which is a beautiful and easy introduction to coding in Python. It is a compact chapter that brings you up to speed with the language with enough information to get you started for the rest of the week. Chapter one is a short introduction to modern machine learning. Chapter two is in my opinion the most important chapter in this course. It is one of the basics of the mathematics used in deep learning. There is no need to panic though. It is really a review of a small section of high school mathematics together with the basics of vectors and matrices. You don't have to solve homework problems or pass a test but at least understanding the basic mathematics will leave you much richer for the experience. It is much better to spend some time with these basic concepts than to be a user of deep learning and not knowing how it really works. I have found that path to be very poor and a known reason why people do not progress in the field. Chapter three puts our understanding of the basics to work. I show you how data flows through a neural network using the simple calculations from chapter two. Chapter four shows you how a neural network learns from the data. It is just an application of the high school mathematics I talked about before. When we create some actual neural networks later in the course the calculations happen behind the scenes and only requires a few lines of code. I reiterate though that understanding what really happens during the learning phase is a true enriching experience. Chapter five is a demo of the flow of data through a neural network and then the learning phase. After this chapter you will have a good understanding of neural networks. In chapter six I show you how to generate simulated data. When you start designing your own neural networks you want as much data as possible. It is very useful when you can simulate data as you gain experience in the design of neural networks and you can save a lot of time when trying to find and clean up external sources of data. Now in chapter seven we create our first proper neural network and solve a classification problem and in chapter eight we create a neural network to solve a regression problem. Chapters nine and ten are suggested reading only and I do want you to read through the material though. There are some more complicated mathematics in chapter ten. Now we will implement the mathematics with single lines of code so don't expect you to understand all the math. Reading through the chapter will give you an understanding of why we include these lines of code in our neural networks. In chapter eleven we import some real data and create a neural network that predicts the presence of heart disease. In chapter twelve I talk about another type of neural network called a convolutional neural network. Now these are great for learning from images. In the final chapter we create and train a network to distinguish between benign and malignant skin lesions on images. And that is where I leave you ready to start your own journey. Remember to read the description down below that will give you an update of which chapters are covered on which days of the course. Before you start on this journey let's go inside so that I can show you how to prepare for this course. We are indoors sitting comfortably at our desk behind our computers and I've opened Google Chrome. Of course we are going to make use of Google Colab as I said and Google Colab is an application just like Google Docs or Google Sheets if you are familiar with those. And the best experience that you are going to have is if you install Google Chrome. So all the links will be available but you can just search for how to install Google Chrome and this is the page that you most probably are going to land up on and you can simply download Chrome from here run the installation irrespective of your operating system and you will have Google Chrome. Once you have Google Chrome of course it's important that you have a Google account. I think the majority of us will have a Gmail account but in case you do not just go to accounts.google.com it will bring you to this sign up page and you can sign up for a Gmail account. Once you have a Gmail account and your Google Chrome is up and running right at the top here you will see this little nine dots Google Apps and if you open that in Google Chrome all your apps are available there. There's Gmail, YouTubers there, the Play Store etc. But somewhere along all of these you'll find your Google Drive. Google Drive is a cloud service. It is much like the hard drive or solid state drive in your computer. Just sits in the cloud and you can upload and create files and download files from that drive as is necessary. So just click on that drive and that is going to bring you into Google Drive. And this is what my Google Drive looks like so it is quite full at the moment and you can see here's my courses for Stellenbosch University and here's the Stellenbosch University subfolder or folder inside of my Google Drive so when you open up your Google Drive you are going to be presented with an empty page because you have not uploaded or created anything and it's very intuitive to navigate. There's a big new button there. You can create a new folder just as you do on your own hard drive. Upload a file or upload a whole folder which is of course what you'll have to do because if you sign up for the course you're going to get an archived file, a zip file that you will have to unzip and that will give you one big folder containing all the files that we're going to use in the course and you're simply going to upload that folder by coming here, clicking upload folder and navigating to that extracted folder and simply upload it. Of course, if you are familiar with your Google Drive you know Google Docs, Google Sheets, Google Slides that would be the same as Microsoft Word, Microsoft Excel, Microsoft PowerPoint, et cetera but if you go down to the bottom you'll see there Google Colab and those are the files that we're going to be working with. In case you do not see Colab available immediately you can come to this website Colab colab.research.google.com and it should bring you to this page and you can click on any of these files here in example and that'll open a Colab file for you which we can see in the background and all you have to do is click this little button here Copy to Drive that's going to make a copy of this Welcome to Colab file onto your Google Drive it'll be available there and once you do that should you go back to your Google Drive, the file will be available here and you should also then be able to say New, More and Colab should appear for you there so just in case it's not there when you start off immediately just follow those steps of going to colab.research.google.com and hit this Copy to Drive button it'll copy this file from Google Systems to your Google Drive We are also going to make use of Kaggle Please read up on that, it's Kaggle.com as you can see, K-A-G-G-L-E and on the last step of the course we are going to use Kaggle just to create a notebook that looks pretty much like a Google Colab file and as much as they're all built on a technology called the Jupyter Notebooks so it's very easy to do, just sign up on the top right there, sign up for free account and on the last day we're going to create Python files inside of Kaggle and Kaggle is going to give us access to a bunch of data it's really something that you should look into it's going to give us access to a bunch of data specifically a bunch of images and we are going to build a deep neural network that can diagnose disease from images and that's it, I look forward to meeting you on the course