Python Machine Learning Tutorial #1 - What is Machine Learning?

Key Takeaways

but what is going on guys welcome to the spite in the Tori series for machine learning and I was a little bit tempted to say data science because we're now starting out with something new the tutorial series on machine learning making machines learn teaching a computer how to learn based on data this is what we're going to do in this series and today's video is the introduction episode so today we're going to talk about the basics and fundamentals of machine learning questions like what is machine learning what types of machine learning are there and what libraries are we going to need for this tutorial series so let us get into the explanations now as we all know I am a professional content producer and that's the reason why I only use professional tools to illustrate my concepts and therefore I use paint today because paint is a very professional tool to do that no seriously I'm going to use paint tree because I don't see the value of creating a fancy PowerPoint presentation around machine learning what it isn't and all that so I'm going to use paint because I think I can explain it better in paint and you'll learn much more in paint and it's also more authentic to me so let us start by defining what machine learning is not and machine learning is not a machine learning from a human even though this might seem that it might seem that this is the case but when a human an expert for example a medical expert or a physicist or a finance expert tells a computer what to do in specific cases this is not machine learning this is not artificial intelligent as a result of that of course the computer may be high behave intelligently so if this is a stockbroker and he tells the computer what to do in which situation and the computer then thus that we might consider that intelligent action but in fact it's not artificial intelligent because the intelligent action was just the expert human expert telling the computer what to do so this is not machine learning this is not what we understand when we talk about machine learning machine learning is having data this should represent data I know it looks ugly but let's ignore that for a second we have some kind of data and we just take that day and feed it into the machine into the machine learning model and then as a result of that the computer or the machine knows what to do so an example for that would be to just read in a couple of CAD images so images of a CAD oh god that looks terrible and also images of dogs for example I'm not going to draw a dot right now but just taking different images of cats and dogs and putting them into the machine learning model and of course we always say okay this one is a cat the next one is also a cat the next one is a dog so that the computer recognizes what does a cat look like and what does a dog look like and then only as a result of that we then feed in some images that the computer has never seen before and in the end it tells us ok this is a cat this is a dog this is a cat and so on so basically what we're doing is we're defining a mathematical machine learning model putting in some so-called training examples so pictures of cats and dogs that are already classified as cats and dogs putting them into the machine learning model and as a result of that this model shall be able to classify unknown images of new cats and dogs so a human there there's no human telling the computer what to do here it's no there's no human that tells the computer a cat is a animal that has pointy ears or something like that the computer figures these out are these features out by himself so this is basically what machine learning is so now that we roughly know what machine learning is let's talk about the different types of machine learning and I'm going to write them down for you here and I hope that does not look too ugly and we're going to talk about the three major types which are supervised learning so supervised learning unsupervised learning and the so called reinforcement learning so these are the three major types of machine learning that we have and we're going to talk about all of them right now so let us start with the left one D supervised learning now a classic supervised learning example would be the one with the cats and the dogs whenever we have a machine learning model and I'm going to use this computer here as a model so this represents the machine learning model whenever I have sorry whenever I have data that is already classified or data that already contains the solution or the desired outcome it is a supervised learning example so it is called supervised learning because we're giving supervision to the model we're telling him or we're telling the model this is a cat this is a dog this is a dog this is a cat and so on so we're supervising him it sorry I'm always referring to the computer as a masculine one we're always giving supervision to the computer by saying look what you see here is a dog or what you see here if we take a different example of handwritten digits what you see here is an eight what you see here is a seven what you see here is a three and so on so we're giving the supervision to the model so that it figures out what to look at and this is supervised learning basically just telling them all of what to do and then of course expecting the results so when we have images of cats and dogs so let's say these are okay oh my god that looks ugly cats and dogs if we feed these images with the classifications in and then we feed some new images in what we want is we want also a result that says cat or doc now with unsupervised learning this is a little bit different with with unsupervised learning we provide data that is not classified for example we could look at a coordinate system and have the heights and let's say this is the height and the weight of people and now what we could do is we could just feed in all the data points and just say these data points could be I'm going to use a different color right now green these data points can be everywhere so we can have some here we can have some here we can have some here and so on so these data points are just there and they don't they are not classified so we don't say these are the overweight people these are the skinny people and so on we just say these are the people and we're now telling our model find some clusters for example or find some patterns and the model will see okay here I see a cluster I cannot tell you that these are the overweight people but they these people are one group one cluster and this is another one and this is another one if you allow for three clusters for example if you say okay only I only allow for two clusters you could say okay we have these and these for example okay not exactly these here so this would be unsupervised learning just providing the examples and of course as a result we don't get labels like cat or dog or overweight or something like that we just scat cluster one cluster two and of course if I have a new data point here let's pick an orange one and it's placed here and I have the class clusters from before so this is one cluster of course I know that this data point also belongs to this cluster but I don't know that this cluster is for example the cluster for overweight people I can just figure that out myself as a human but the model doesn't tell me that if we take the classification example of supervised learning in this case what I would do is I would say okay these are all green dots went while training the example this is not something to model this this is something that I do I give this example to the model and I just say these are the overweight people for example they have a lot of weight and a very small height so these are the overweight people they have they're all red for example and I say that to the model so I'm supervising them all and telling him that these are or telling it sorry that these are the red the red people the overweight people then I could say we have the blue people that are the skinny people so I can say okay this is blue and then we have the small and skinny people the Greenpoint's and then I have a new data point let's pick I don't know gray and then I have a new data point and the model can then tell me that this is an overweight person this would be the supervised example but if I don't use these classifications it's just unsupervised and I look for patterns so now let us get to the last one which is reinforcement learning and reinforcement learning is a entirely different type of machine learning reinforcement learning works with agents and environments for example what I could have is an environment like this just a ground and my agent is a person and the goal for that person is to jump as high as possible this is a classic example of that and of course this agent is a computer or some programmed thing some programmed entity and it has control over certain things for example it can bend its knees and it can crouch it can try to jump it can do different things but we don't tell this entity what to do to perform an action we just rewarded for doing the right action so basically we start with random action maybe this person will walk left and right enough and something like that and what we do here is we say okay for walking left and right you get zero points or something like that if you crouch onto the ground or if you go down onto the ground you'll receive minus two points because we want you to move up so as soon as the person jumps a little bit we say okay this is a score of 0.5 points for example and if this person jumps pretty high we can say okay this is a score of 1.75 points and because of these rewards and punishments this agent learns what to do so we're rewarded for jumping so it will always adapt and train itself so that it jumps higher so this is reinforcement learning has a lot to do with genetic algorithms and evolution so this is a very interesting field but these are basically the three different types of machine learning so last but not least let's talk about the libraries that we're going to need this tutorial series and in this video we're going to install two of them scikit-learn and tensorflow we're going to use these two libraries for this tutorial series scikit-learn is for all the traditional machine learning algorithms like basic key neighbors classification regression support vector machines clustering and all that so these are the basic fundamentals and traditional elements of machine learning and then we have tensorflow for deep learning and neural networks later on so we're going to install both of them and we're going to run CMD for that and activate the Anaconda environment if you don't have one but you want one you can check out my data science tutorial series there I explained how to set up an account properly so now I activated my environment and I'm going to type pip install scikit-learn this is how you install so I could learn so you just this is the library as I said for the traditional machine learning algorithms regression cluster II and classification and so on and I'm going to get back to you as soon as the installation is finished all right there we go and now we're going to install tension flow as well notice however that tension flow sometimes produces problems if you're using the latest version of Python because sometimes when a new python version is released tensorflow is not yet compatible with it so maybe turn to flow is not available if you're using the latest version so you maybe want to downgrade the version to the 2nd latest but if it works you just type it install tensor flow to install tensorflow on your machine and again I'm getting back to you after the installation alright the installation is done and we can now proceed with the coding in pycharm we're going to use Python for this tutorial series as well of course you can choose any development environment that you want or that you like you can use spider you can use Jupiter notebook whatever you want but I'm going to use PyCharm for this tutorial series as well and now we have installed all the packages that we're going to need so that's it introduction into the machine learning series with Python in the next video we're going to start out with some coding I think we're going to start with linear regression let's see but definitely we're not going to do more Theory we're going to get into the coding in the next episode so I hope you enjoyed this episode I hope you learned something if so hit the like button to support this channel also feel free to ask questions and get feedback in the comment section down below and of course subscribe to this channel if you want to see more so thank you very much for watching see you in the next video and bye [Music]