Deep Learning Specialization
Deep Learning Specialization
What is Deep Learning?
Definition of Deep Learning:
Deep Learning. Ever wondered how google translates an entire webpage to a different language in a matter of seconds or your phone gallery groups images based on their location all of this is a product of deep learning.
But what exactly is deep learning deep learning is a subset of machine learning which in turn is a subset of artificial intelligence artificial intelligence is a technique that enables a machine to mimic human behavior machine learning is a technique to achieve AI through algorithms trained with data.
And finally deep learning is a type of machine learning inspired by the structure of the human brain in terms of deep learning this structure is called an artificial neural network let's understand deep learning better and how it's different from machine learning say we create a machine that could differentiate between tomatoes and cherries if done using machine learning.
We'd have to tell the Machine the highlights dependent on which the two can be separated these highlights could be the size and the sort of stem on them with deep learning then again the highlights are selected by the neural system without human mediation obviously that sort of freedom comes at the expense of having an a lot higher volume of data to train our machine.
Now let's dive into the working of neural networks here we have three students each of them write down the digit nine on a bit of paper strikingly they don't all compose it indistinguishably the human mind can without much of a stretch perceive the digits however imagine a scenario in which a PC needed to remember them that is the place deep learning comes in here's a neural system prepared to distinguish written by hand digits each number is available as a picture of multiple times 28 pixels.
Now that adds up to an aggregate of 784 pixels neurons the center substance of a neural system is the place the data handling happens every one of the 784 pixels is taken care of to a neuron in the primary layer of our neural system this structures the information layer on the opposite end we have the yield layer with every neuron speaking to a digit with the shrouded layers existing between them the data is trans starting with one layer then onto the next over interfacing channels each of these has a worth joined to it.
And thus is known as a weighted Channel all neurons have an extraordinary number related with it called predisposition this inclination is added to the weighted total of sources of info arriving at the neuron which is then applied to a capacity known as the actuation work the consequence of the enactment work decides whether the neuron gets initiated each actuated neuron gives data to the accompanying layers this proceeds up till the subsequent last layer the one neuron initiated in the yield layer compares to the info digit the loads and predisposition are persistently acclimated to deliver a well-trained network.
So where is deep learning applied in customer support when most people converse with customer support agents the conversation seems so real they don't even realize that it's all things considered a bot on the opposite side in clinical consideration neural systems recognize malignant growth cells and dissect MRI pictures to give definite outcomes self-driving vehicles.
What appear as though sci-fi is presently a reality Apple Tesla and Nissan are just a couple of the organizations chipping away at self-driving vehicles so deep learning has a tremendous extension yet it also faces a few restrictions the first as we talked about before is information while deep learning is the most productive approach to manage unstructured information a neural system requires an enormous volume of information to Train we should expect we generally approach the essential measure of information preparing this isn't inside the capacity of each machine.
And that carries us to our second confinement computational force preparing and neural system requires graphical handling units which have a huge number obviously when contrasted with CPUs and GPUs are obviously progressively costly and finally we come down to training time deep neural networks take hours or even months to train the time increases with the amount of data and number of layers in the network so here a short quiz for you arrange the following statements in order to describe the working of a neural network.
The bias is added be the weighted sum of the inputs is calculated see specific neuron is activated D the result is fed to an activation function leave your answers in the comments section below three of you stand a chance to win Amazon vouchers so hurry some of the popular deep learning frameworks include tensorflow high torch caris deep learning Forge a cafe.
And Microsoft cognitive toolkit considering the future predictions for deep learning and AI we seem to have only scratched the surface in fact horse technology is working on a device for the visually impaired that utilizes deep learning with PC vision to portray the world to the clients recreating the human psyche at the whole might be not only a scene of sci-fi or too long what's to come is for sure brimming with shocks and that is deep learning for you in short.
Deep Learning vs Machine Learning:
Deep learning vs Machine Learning. The best way to explain this is with an example so you probably come across an app that allows you to take a picture of your face and test our different type of makeup items or you can take a picture of someone and then it helps you find similar clothes.
So the first step to any of these apps is to recognize what the app or the product or the artificial intelligence is actually looking at so is this a dress is a pair of pants is the jacket is a person's face what is it so in this example we're gonna look at like how machine learning and deep learning is used to enable your app or your AI to understand what is it it is looking at so in this article we're gonna cover through what is machine learning what is deep learning why all of a sudden it is everywhere and how do you pick one over the other so on a broader spectrum.
If you think of artificial intelligence as the main umbrella machine learning is a subset of artificial intelligence that's how AI is has been happening in last 20 years but deep learning is a new subset of machine learning that has come in to exist and then just pretty recently and it has revolutionized and changed the way artificial intelligence works so what is machine learning so algorithms that give the ability to computers to learn from data and make predictions so what does that mean so going back to a previous example let's say give an app and you're used to take a picture.
How will an app tell or artificial intelligence tell that is it looking at a mini-dress versus a pair of pants and we're gonna use this as an example so what do you do in machine learning is you when the pictures are uploaded you define what is the definition of a dress so basically you will tell the computer that a dress has a different type of neck size it has shoulder traps it probably doesn't have any zippers it has holes for arms on each side of the neck so you're basically defining the features of what a dress looks like and the same thing you will do for what a pant look like.
So it has a split for the legs it is opening on the top which is size can vary from in diameter from you know 24 inches to 45 inches you know it possibly has buttons it has belt buckles or it has can have a belt it has pockets on the front so when we as humans look at a pair of pants how do we tell that it's actually a pair of pants we look at these things so you are basically defining this for the computer so computer and artificial intelligence can actually create a model of what is the definition of features of a pair of pants or a pair of dress s and this is the way you can define the features you can tell you can label different parts you know you can have the distance between the shoulder could be from here to here things.
Like that so once it is done so the typical flow from machine learning goes into you upload images or data could be sound could be anything you define what those features are then you create a model and you can have hundreds of models depending on which what your part of this product is so if this is a fashion app you know you have jackets eve pants your swimsuits your shorts.
So there are like hundreds and hundreds of items so each one of that item can actually have its own model which defines these features so you can have multiple ones and now the machine learning algorithm is ready to make a prediction so when you upload a next picture or expose it to the next item.
It looks at all those models and try to define what is it actually looking at and this is how the machine learning part actually makes predictions so for example if you showed her a picture of this image if you look at the image and look at the model it created based on the features you told her that this is actually this is a hot dress is supposed to look like and it'll give you hey yeah this is a dress and if you showed us image like this.
If it doesn't follow the typical model or the features that you define for a dress it's going to give you answer no this is not a dress so this is how machine learning works and you have been seeing examples of these for you know years so spam email for example is is actually a type of machine learning.
Because you define what is the definition of a spam email so for Google for example you will tell hey any email that has Nigerian prints on it and it has some viagra in it or it has multiple lengths of multiple images or the sender's name or the sender's email is not defined so you're defining the features what is the definition of expand email and when the model looks at hey this is this checks are like 80% or 90% of the features that that's I would think that this defines a spam email then it will start putting it into your spam folder same thing went with YouTube video recommendation.
That it looks at which videos are you watching which was the author how long is the video what is the video again which topic the video is on you notice the video has sound versus just instrumental with your just music and what kind of music so this is what YouTube was taking into consideration and based on that it will recommend the next video to you so machine learning had existed for a long time and this is how it has been working you probably without even realizing it.
So now let's get to our next section which is what is deep learning so as we talked about the deep learning is a subset of machine learning so if you want to define deep learning as deep learning is a subfield of machine learning concerned with algorithms inspired by structure and function of the brain called artificial neural networks so if you look at our our brain we have billions of neurons and every time you look at something or you learn a new piece of information that neurons the electricity is firing between neurons and that's our it establishing connections or in the future when you come across a similar looking image or sound or whatever information that could be it looks at those connections and helps you define what you are looking at so.
So with deep learning is it based on very similar things or deep learning what scientists have created is called artificial neural network which is a layer of neurons that piece of information goes through and then it defines what it is so we'll look into it what exactly but that's the basic of composition so as we looked at for machine learning we looked at in the past you know typical model you upload the pictures you manually label those features and then it predicts with what it is so manually labeling those features.
And definition is still prone to human error because if you it will take a lot of time and if you're not able to define it correctly it could be an issue so what the best benefit of deep learning is you don't have to do this manual labeling of features anymore so what deep learning does is its neural network automatically creates its own model and automatically defines what if things are the features or the model or the definition of a dress or a pair of pants s and then it'll learns on its own.
So what you are doing is you are actually just uploading the images and you're telling it this is an image of a dress but you are not telling what is a feature of a dress and you upload thousands of images of dress in different conditions you know from different poses like side pose front post back pose and based on that it automatically creates what is the definition of a dress or what are the features of the dress.
And then it creates a model on its own and this is how the neural network basically works like that you upload images and these are the layers of different neurons through which the information flows and with each subset of these layers it creates a different definition or a model of what is it looking at.
And in the end it gives you an output and defines what it's looking at so why is deep learning everywhere all of a sudden so in order to do deep learning you basically need three things you need lots of label data as we talked about in the beginning that when you are uploading images of dresses you have to tell it this is the image of a dress or this is image of a pan.
So all this data needs to be labeled and a second thing you need you need high performance GPUs which basically can process this tons and tons of information at a faster rate and third you need sophisticated algorithms to help you define and make sense of what is looking at so what did happen in the past was we did not have enough label of data and the profile performance GPUs were extremely so this is a fast correlation of what machine learning and deep learning is and in the event that you have any inquiries kindly leave them in the discussion for input and I'll gladly reply.
Is Deep Learning AI?
Is Deep Learning AI? Now my topic of discussion is AI vs Machine Learning vs Deep Learning. These are the term which have confused a lot of people and if you too are one among them, let me resolve it for you.Well artificial intelligence is a broader umbrella under which machine learning and deep learning come you can also see in the diagram that even deep learning is a subset of machine learning so you can say that all three of them the AI the machine learning and deep learning are just the subset of each other.
So let's move on and understand how exactly the differ from each other. So let's start with artificial intelligence. The term artificial intelligence was first instituted in the year 1956. The concept is pretty old, but it has gained its popularity recently.
But why well, the reason is earlier we had very small amount of data the data we had Was not enough to predict the accurate result, but now there's a tremendous increase in the amount of data statistics suggest that by 2020 the accumulated volume of data will increase from 4.4 zettabyte stew roughly around 44 zettabytes or 44 trillion GBs of data along with such enormous amount of data.
Now, we have more advanced algorithm and high-end computing power and storage that can deal with such large amount of data as a result. It is expected that 70% of Enterprise will Implement ai over the next 12 months which is up from 40 percent in 2016 and 51 percent in 2017.
Just for your understanding what does AI well, it's nothing but a technique that enables the machine to act like humans by replicating the behavior and nature with AI it is possible for machine to learn from the experience. The machines are just the responses based on new input there by performing human-like tasks.
Artificial intelligence can be trained to accomplish specific tasks by processing large amount of data and recognizing pattern in them. You can consider that building an artificial intelligence is like Building a Church, the first church took generations to finish. So most of the workers were working in it never saw the final outcome those working on it took pride in their craft building bricks and chiseling stone that was going to be placed into the great structure.
So as AI researchers, we should think of ourselves as humble brick makers whose job is to study how to build components example Parts is planners or learning algorithm or accept anything that someday someone and somewhere will integrate into the intelligent systems some of the examples of artificial intelligence from our day-to-day life our Apple series just playing computer Tesla self-driving car and many more these examples are based on deep learning and natural language processing.
Well, this was about what is AI and how it gains its hype. So moving on ahead. Let's discuss about machine learning and see what it is and white pros of an introduced. Well Machine learning came into existence in the late 80s and the early 90s, but what were the issues with the people which made the machine learning come into existence? Let us discuss them one by one in the field of Statistics.
The problem was how to efficiently train large complex model in the field of computer science and artificial intelligence. The problem was how to train more robust version of AI system while in the case of Neuroscience problem faced by the researchers was how to design operation model of the brain.
So these are some of the issues which had the largest influence and led to the existence of the machine learning. Now this machine learning shifted its focus from the symbolic approaches. It had inherited from the AI and move towards the methods and model. It had borrowed from statistics and probability Theory.
So let's proceed and see what exactly is machine learning. Well Machine learning is a subset of AI which The computer to act and make data-driven decisions to carry out a certain task. These programs are algorithms are designed in a way that they can learn and improve over time when exposed to new data. Let's see an example of machine learning. Let's say you want to create a system which tells the expected weight of a person based on its side.
The first thing you do is you collect the data. Let's see there is how your data looks like now each point on the graph represent one data point to start with we can draw a simple line to predict the weight based on the height.
For example, a simple line W equal x minus hundred where W is waiting kgs and edges hide and centimeter this line can help us to make the prediction. Our main goal is to reduce the difference between the estimated value and the actual value. So as to accomplish it we attempt to draw a straight line that fits through all these various focuses and limit the mistake.
So our main goal is to minimize the error and make them as small as possible decreasing the error or the difference between In the actual value and estimated value increases the performance of the model further on the more data points. Well, this was a basic discussion in case you have any doubt feel free to add your query to the comment section.
Reinforcement Deep Learning:
Reinforcement Deep Learning. From the astonishing outcomes and vintage Atari games deep Minds triumph with alphago shocking forward leaps in mechanical arm control and in any event, beating proficient players at 1v1 dota the field of reinforcement learning has actually detonated as of late since the time the amazing achievement on the imagenet arrangement challenge in 2012 .
The accomplishments of managed deep learning have kept on accumulating and individuals from a wide range of foundations have begun utilizing deep neural nets to tackle a wide scope of new assignments remembering how to learn savvy conduct for complex unique conditions so in this scene .
I will give a general presentation into the field of reinforcement learning just as an outline of the most testing issues that we're confronting today in case you're searching for a strong presentation into the field of deep reinforcement learning then this article is actually what you're searching for.
Thus, in its 2017 Peter Emil gave an exceptionally rousing demo before a huge crowd of the absolute most brilliant personalities in AI and machine learning so you indicated this video where a robot is tidying up a lounge room presenting to someone a jug of brew and fundamentally doing an entire scope of everyday errands that robots in science fiction motion pictures can manage beyond a shadow of a doubt and afterward toward the finish of the video dwindle uncovered that the robots activities were quite remote-constrained by a human administrator and the takeaway from this demo.
I believe is a significant one it essentially says that the robots we've been working throughout recent decades are truly flawlessly equipped for doing a wide scope of helpful undertakings yet the issue is that we can't insert them with the required intelligence to do those things so fundamentally making valuable best in class mechanical autonomy is a product challenge and not an equipment issue thus for reasons unknown, having a robot figure out.
How to accomplish something straightforward like getting a jug of lager can be a difficult assignment thus in this article I need to acquaint you folks with the entire subfield in machine learning that is called reinforcement learning which I believe is one of the most encouraging headings to really get to smart automated conduct so in the most well-known machine learning applications individuals use what we consider directed learning and this implies you give a contribution to your neural system model.
Yet you know the yield that your model should deliver and subsequently you can register slopes utilizing something like the back proliferation calculation to prepare that system to create your yields.
So envision you need to prepare a neural system to play the round of pong what you would do in an administered setting is you would have a decent human gamer play the round of pong for several hours and you would make an informational index where you log the entirety of the casings that human is seeing on the screen just as the moves that he makes in light of those edges so whatever is pushing the up bolt or the down bolt and we would then be able to take care of those info outlines through a basic neural system that at the yield can deliver two basic activities.
It's either going to choose the up activity or the down activity and by basically preparing on the informational index of the human ongoing interaction utilizing something like back engendering we can really prepare that neural system to duplicate the activities of the human gamer yet there are two huge drawbacks to this methodology so from one viewpoint in the event.
That you need to do regulated learning you need to make an informational collection to prepare on which isn't constantly a simple activity and then again on the off chance that you train your neural system model to just copy the activities of the human player well then by definition your specialist can never be better at playing the round of pong than that human gamer for instance on the off chance that you need to prepare a neural net to be better at playing the round of gold and the best human then by definition we can't utilize managed learning.
So is there an approach to have an operator figure out how to play a game altogether without anyone else well luckily there is and this is called reinforcement learning so the structure and reinforcement learning is very like the typical system in managed learning so we despite everything have an information outline we run it through some neural system model and the system delivers a yield activity we either up or down.
However the main distinction here is that now we don't really have the foggiest idea about the objective name so we don't know in any circumstance whether we ought to have gone up or down in light of the fact that we don't have an informational collection to prepare on and in reinforcement learning the system that changes input edges to yield activities is known as the strategy Network.
Now perhaps the least difficult approaches to prepare an arrangement organize is a technique called strategy angles so the methodology in arrangement inclinations is that you begin with a totally irregular system you feed that arrange an edge from the game motor it creates an arbitrary up with activity you know either up or down you send that activity back to the game motor and the game motor creates the following casing and this is the means by which the circle proceeds and the system for this situation.
It could be a completely associated organize yet you can clearly apply convolutions there too and now as a general rule the yield of your system is going to comprise of two numbers the likelihood of going up and the likelihood of going down and what you will do while preparing is really test from the circulation.
So you're not continually going to rehash the equivalent precise activities and this will permit your operator to kind of investigate the condition a piece haphazardly and ideally find better rewards and better conduct now significantly in light of the fact that we need to empower our specialist to adapt totally without anyone else the main input that we're going to give it is the scoreboard in the game.
At whatever point our operator figures out how to score an objective it will get a compensation of +1 and on the off chance that the rival scored an objective, at that point our operator will get a punishment of less 1 and the whole objective of the operator is to upgrade its strategy to get however much prize as could reasonably be expected so as to prepare our arrangement organize the principal thing.
We're going to do is gather a lot of understanding so you're simply going to run an entire bundle of those game casings through your system select irregular activities feed them once again into the motor and simply make an entire pack of arbitrary pong games and now clearly since our specialist hasn't got the hang of anything helpful yet it's going to lose the vast majority of those games yet indeed some of the time our operator may get fortunate here and there it's going to arbitrarily choose an entire grouping of activities that really lead to scoring an objective.
Deep Learning and neural networks:
Deep Learning and neural networks. Last summer my family night visited Russia even though none of us go to read Russian we did not have any trouble in figuring our way out all thanks to Google's real-time translation of Russian boards into English this is just one of the several applications of neural networks neural networks.
Form the base of deep learning a subfield of machine learning where the algorithms are inspired by the structure of the human brain neural networks take in data train themselves to recognize the patterns in this data and then predict the outputs for a new set of similar data.
Let's understand how this is done let's construct a neural network that differentiates between a square circle and triangle neural networks are made up of layers of neurons these neurons are the core processing units of the network first we have the input layer which receives the input the output layer predicts our final output in between exist the hidden layers which perform most of the computations required by our network here's an image of a circle this image is composed of 28 by 28 pixels.
Which make up for 784 pixels each pixel is fed as input to each neuron of the first layer neurons of one layer are connected to neurons of the next layer through channels each of these channels is assigned a numerical value known as weight the inputs are multiplied to the corresponding weights and their sum is sent as input to the neurons in the hidden layer each of these neurons is associated with a numerical value called the bias.
Which is then added to the input sum this value is then passed through a threshold function called the activation function the result of the activation function determines if the particular neuron will get activated or not an activated neuron transmits data to the neurons of the next layer over the channels in this manner the data is propagated through the network this is called forward propagation in the output layer.
The neuron with the highest value fires and determines the output the values are basically a probable for example here are near unassociated with square has the highest probability hence that's the output predicted by the neural network of course just by a look at it we know our neural network has made a wrong prediction but how does the network figure this out note that our network is yet to be trained during this training process along with the input our network.
Also as the output fed to it the predicted output is compared against the actual output to realize the error in prediction the magnitude of the error indicates how wrong we are in the sign suggests if our predicted values are higher or lower than expected the arrows here give an indication of the direction and magnitude of change to reduce the error this information is then transferred backward through our network this is known as back propagation.
Now based on this information the weights are adjusted this cycle of forward propagation and back propagation is iteratively performed with multiple inputs this process continues until our weights are assigned such that the network can predict the shapes correctly in most of the cases this brings our training process to an end you might wonder how long this training process takes honestly neural networks may take hours or even months to train.
But time is a reasonable trade-off when compared to its scope let us look at some of the prime applications of neural networks facial recognition cameras on smartphones these days can estimate the age of the person based on their facial features this is neural networks at play first differentiating the face from the background.
And then correlating the lines and spots on your face to a possible age forecasting neural networks are trained to understand the patterns and detect the possibility of rainfall or arise and stock prices with high accuracy music composition neural networks can even learn patterns and music and train itself enough to compose a fresh tune.
So here's a question for you which of the following statements does not hold true activation functions are threshold functions B error is calculated at each layer of the neural network see both forward and back propagation take place during the training process of a neural network D.
We are still taking baby steps the growth in this field has been foreseen by the big names companies such as Google Amazon and Nvidia have invested in developing products such as libraries predictive models and intuitive GPUs that support the implementation of neural networks the question dividing the visionaries is on the reach of neural networks to what extent can we replicate the human brain we'd have to wait a few more years to give a definite answer.
Deep Learning Applications:
Deep learning applications. we're gonna cover some of the many uses is being applied to healthcare entertainment composing music image coloring robotics image captioning advertising earthquake prediction and there's many more it's just an exciting time to be in the field of AI and deep learning and artificial intelligence now all these fields are just booming it's just amazing what's going on.
So let's take a quick glance at some of the things going on and we'll start with our healthcare we take a look at healthcare deep learning is reshaping healthcare industry by delivering new possibilities to improve people's lives and we have computer aided disease detection I know a number of startups have come up where they're working on you take a picture like if you have some kind of irritation or rash and they use that or even you know some infection in the eye.
And they use the deep learning to help identify what it is hey this is 99.99% nothing don't worry about it or this is 93% chance and usually doesn't come out ninety three percent chance if you do deep learning but it says hey you should probably take this into the doctor you know I'm take a look at it and see what it is these are the things that could be analyzing genomes that's a big one very controversial nowadays with all the different genome editing and options they have and things are doing but deep learning is definitely diving into the genome projects.
We have discovering new drugs and this there's so many directions for new drugs going on it's like one of the booming industries especially in the US but all across the world and they range from exploring different plants and the opening up of let's say marijuana in the US and number cities are exploring that for medical use - how do you reprogram T cells in the human body to combat disease.
There's whole industries and data analysis another one is chemistry how do you use a chemical cell or a chemistry molecule not a chemical molecule to imitate the T cell so you could then put those molecules in there and then they grab the diseased cells out or the cancer cells or whatever out just like the human body would so discovering new drugs and how to apply that is just a huge industry all these are huge industries and medical imaging will even take a glance closer at that.
But being able to analyze you know we get your MRI you get your cat scans the doctors can spend a long time looking at those and taking careful measurements and most of that's done by hand but as we start applying deep learning that deep learning can do a lot of that work not only do it the same but it be uniform from one hospital to the next meaning that as a industry it continues to progress.
And you get better results and better services so deep learning is reshaping healthcare industry by delivering new possibilities to improve people's life and deep learning helps an early detection of cancer cells and tumors improves a time-consuming process of synthesizing new drugs and inverse sophisticated medical instruments and not even sophisticated they're using iPhones in deep learning to help with medical diagnosis so it's at all levels it's just amazing what they're doing in the healthcare industry with deep learning and they've just barely tapped what there's possible they're just now evolving into the beginning of this where it's exploding out there so entertainment that's always a fun one a lot of people go into the entertainment and deep learning.
Because it's just fun you know what's my favorite movie all that fun stuff deep learning is used in the entertainment industry such as Netflix Amazon and filmmaking and you have like your recommendations that's probably the most common one you see and you can see here where they have India transform the true story the third pillar now markets in the state leave.
So it comes up and says hey this is based on what whoever and of course this is showing up on the Amazon or Amazon Prime Netflix does a good job of grouping movies for you it says since you like these movies you might like these movies over here same thing with VEVO it does that also Amazon Netflix and VEVO use recommender systems to provide personalized experience to its viewers using their show preferences time of access history etc so really this is one of the cool things is instead of having to dig for the information.
Deep learning starts doing that for you so instead of spending hours trying to figure out what you watch you can actually just enjoy the movies and have them come up deep learning is used in the entertainment industry such as Netflix Amazon and filmmaking continue on and deep learning.
We have things like the IBM Watson this is just really cool I didn't know this till I read this till we did this slide wimbledon 2018 used IBM watson to analyze player emotions and expressions through hundreds of hours of footage and then it auto generated highlights for the telecast.
So you can imagine somebody sitting there looking through all the boring stuff I want to look at the interesting stuff it helps us sort that out again you know just like the Netflix and referral it helps you find your choices without having to do all the heavy work of digging through really bad berated movies and music and audio generation. So, this is the short review about deep learning applications.
Deep Learning With MATLAB:
Deep Learning With MATLAB. I'm here to reveal to you how MATLAB utilizes clear to make a deep neural system without any preparation our demo has explicit application to picture handling and acknowledgment yet we feel like pictures are entirely simple to identify with.
And it's a genuinely notable utilization of neural systems above all we need to make deep learning open to everybody and you'll have the option to get your hands on everything that we show you and expand on them and begin utilizing your own systems all alone so for those of you who are overly acquainted with preparing systems alongside methods to make them increasingly precise MATLAB will be incredible for you on the grounds that as you would expect we give you the instinctive language structure.
And capacities which will permit you to effectively actualize your assignments for those of you new to the deep learning field and need to consider going all in with this innovation the degree of what you can promptly do may be constrained to picture acknowledgment.
However I'm certain that it'll furnish you with all that anyone could need material to begin and have a great deal of fun with neural systems so this is what we will would we like to prepare a system to perceive four distinct creatures felines hounds frogs and deer to do that we will acquaint pictures of every creature with our system characterize the layers of our system.
And afterward even a solitary line of code advise malla to prepare and make our system without any preparation then we'll try out our system by indicating it new pictures that it hasn't seen previously and check its exactness to set things up we will go into this index and draw 5,000 pictures of every creature and two separate organizers now in case you're crunching the numbers that is 20,000 pictures all out and for those of you who are simply intrigued and attempting to set you may be thinking stand by.
So you anticipate that me should peruse this article and afterward go minister 20,000 pictures before I can even begin well you can in the event that you need or you can do what we took favorable position of work that is as of now been done for this situation we got the entirety of our pictures can be freely accessible see for 10 datasets which truly just includes downloading and removing one major compress record so fortunately setting up this demo is forlorn subject to your system rates.
And processor power individually that being said preparing Network without any preparation requires a considerable amount of information so consistently search for chances to expand on past work like this demo how about we investigate the center code required to execute our preparation you can see this part which determines the creature names and in this part pointing MATLAB to the envelope containing that preparation information and most definitely that is it.
So now we're going to disclose to MATLAB how we need the deep system to be prepared each neural system has a progression of layers and the more layers it has the deeper the system now each layer takes in information from the past layer changes the information and afterward passes it on so the main layer takes in the crude information picture and when we get to the last layer it's going to ideally let out the right name of the creature in the first picture so here are the layers that we've decided to actualize.
Deep Learning CNN:
Deep Learning CNN. In this passage we'll be talking about convolutional neural networks a convolutional neural system otherwise called a CNN or comp net is an artificial neural system that is so far been most prevalently utilized for dissecting pictures despite the fact that picture investigation has been the most boundless utilization of CNN's they can likewise be utilized for other information examination or arrangement issues too most for the most part we can think about a CNN.
As an artificial neural system that has some sort of specialization for having the option to select or identify examples and comprehend them this example location is the thing that makes CNN so helpful for picture examination so if a CNN is only some type of an artificial neural system what separates it from only a standard multi-layer perceptron or MLP well a CNN has concealed layers called convolutional layers and these layers are decisively.
What makes a CNN well a CNN currently CNN's can and for the most part have other non convolutional layers too however the premise of a CNN is the convolutional layers good so what do these convolutional layers do simply like some other layer a convolutional layer gets input at that point changes the contribution to some way and afterward yields the change contribution to the following layer with a convolutional layer this change is a convolution activity.
We'll return to this activity in a piece until further notice how about we take a gander at an elevated level thought of what convolutional layers are doing as referenced before convolutional neural networks can tech examples and pictures all the more correctly the convolutional layers can recognize designs well really how about we be somewhat more exact than that with each convolutional layer we have to determine the quantity of channels the layers ought to have.
And will talk in fact about what a channel is in only a couple of seconds yet for the present comprehend that these channels are really what distinguish the examples now when I state that the channels can distinguish designs what unequivocally do I mean by designs well think about what amount might be going on in any single picture numerous edges shapes surfaces objects and so forth so one kind of statement design that a channel could recognize could be edges and pictures.
so this channel would be called an edge indicator for instance a few channels may distinguish corners some may distinguish circles different squares now these straightforward and sort of geometric channels are what we'd see toward the beginning of our system the deeper our system goes the more advanced.
These channels become so in later layers as opposed to edges in basic shapes our channels might have the option to distinguish explicit articles like eyes ears hair or hide quills scales and snouts even and in significantly deeper layers the channels can take considerably progressively modern items like full pooches felines reptiles and feathered creatures.
To comprehend what's really occurring here with these convolutional layers and their particular channels we should take a gander at a model so state we have a convolutional neural system that is tolerating pictures of written by hand digits like from the pardon ADA set and our system is characterizing them into their individual classifications of whether the pictures of a 1 2 3 and so on.
Deep Learning with TensorFlow:
Deep Learning with TensorFlow. In this paragraph let's have a quick introduction to tensor flow this is probably one of the most popular deep learning frameworks out there right now guys well this session is intended to be useful for anyone who is considering to start learning about tensor flow and need an easy start or if you're considering building a career and deep learning as well.
Let's begin by checking out the agenda for today guys this session is pretty much simple and straightforward we'll be starting out with what tensor flow actually is then we need to find out the easiest way to learn because I'm sure you guys are all excited to get started with tensor flow right.
So then we need to check out on how we could get tensor flow in your machine and get you coding and then we'll talk a little bit about the architecture and the tensor flow pipeline as well so the overall agenda is to get you I started with tensor flow and to make sure that you do it in the easiest way possible well.
Let's begin what is tensor flow well as per the textbook definition it is an open source library which is used for numerical computation and large-scale machine learning well tensor flow is beyond all that and it is an integral part of almost everything that involves machine learning or deep learning it definitely is the brainchild of all the awesome people at Google and it's integrated with all of their apps as well well I personally love anything open-source it opens up so much opportunities for development and improvement.
And all right so it is available on desktop and mobile platforms as well which is definitely a big positive and brings in more people and well this leads to one of the largest machine learning community coming together under one roof so yes tensorflow has the largest community of learners and collaborators when compared to all the other deep learning frameworks right now and majority of the big firms everything from social media all the way till an airline company make use of tensor flow.
So tensor flow definitely has some amazing industry to action so let's get started offensive flow is getting all this traction I am sure you guys are curious about why it became so popular right well I'm gonna pick one of the amazing use cases that I personally used when I was on a tour so consider this guys you and your family are going to Japan for a vacation well majority of the sign boats are in nice and most of the people speak Japanese only well this can be a problem as it leads to a huge communication gap.
Well let's see if tensorflow can help us here somehow well with image recognition systems integrated into google applications it makes it so seamless that I was blown away when I use it for the first time right so you have Japanese board in front of you which makes absolutely no sense to you because you don't know Japanese right well easy open up the Google Translate camera and just point it at the board and that's it the entire text is translated for you in real time well this is one among the major reasons there are millions of happy travelers who don't worry about their language barriers.
Well so basically since its open-source it is free and it makes sure that you don't get cheated upon in a foreign land as well and what's even better is that you can convert entire documents at a single click as well well guys this is just one of the amazing use cases offensive flow and there are thousands more and the best part about it is that it's really easy to get started with tensorflow as well to be honest with you.
There are many easy ways to get started with tensorflow in my opinion Google collaboratory is by far the easiest to get started you might be wondering what collaboratory is right well it's really simple it's basically a Jupiter notebook hosted on the Google cloud platform it automatically allocates memory.
And provides you with all the tools that you need to get started however for complex computations you will require the use of a GPU right well Google has got us covered there as well fast computations can now be done because collaboratory allocates a GPU for us as well.
All of this is done free and it's real cool but there is another traditional approach that many of you guys might prefer right which is definitely installing pycharm and then installing all the modules inside it as well but I prefer Google collab just for its simplicity and ease of use plus your code always stays on the cloud.
Basics of Deep Learning
Conclusion:
So companions in this article I was given a detail presentation about Deep Learning, and I answer few questions about Deep Learning that people asked from me.
Expectation you make the most of my detail data about Deep Learning! And if you had any questions about Deep Learning leave a remark I will answer.
What's more, If you delighted in it!
If it's not too much trouble leave a remark!
What's more, I trust you get familiar with Deep Learning in the course interface given underneath!
In this course, you will get familiar with the establishments of Deep Learning, see how to fabricate neural systems, and figure out how to lead fruitful AI ventures. You will find out about Convolutional systems, RNNs, LSTM, Adam, Dropout, BatchNorm, Xavier/He instatement, and that's only the tip of the iceberg. You will take a shot at contextual investigations from human services, independent driving, gesture based communication perusing, music age, and common language handling. You will ace the hypothesis, yet additionally perceive how it is applied in industry. You will rehearse every one of these thoughts in Python and in TensorFlow, which we will educate.
You will likewise get notification from many top pioneers in Deep Learning, who will impart to you their own accounts and offer you vocation guidance.
Computer based intelligence is changing different enterprises. In the wake of completing this specialization, you will probably discover inventive approaches to apply it to your work.
We will assist you with acing Deep Learning, see how to apply it, and manufacture a profession in AI.
So Hurry up! Select this course to become familiar with Deep Learning!
Much obliged to you!
0 Comments
Please do not enter any spam link in the comment box.