TensorFlow | What Is TensorFlow | How TensorFlowWorks | TensorFlow Explained | Intellipaat

[Music] hello everyone and welcome to today's session on what is tensorflow tensorflow is an open source toolkit for numerical computation and large scale machine learning that was created by the google ring team and was first released in the public in 2050. tensorflow combines a variety of machine learning and deep learning models and also algorithm known as neural networks that makes usable using conventional programming idioms it provides an easy front-end api for constructing apps using python or javascript while running those applications in high performance c plus plus before knowing more about what is tensorflow please do not forget to hit the subscribe button and click the bell icon so without further ado let's check today's agenda we are going to start with what is tensorflow then we are going to discuss about architecture of tensorflow later on we are going to discuss about tensorflow components next we are going to discuss where can tensorflow run moving ahead we are going to discuss about algorithms for tensorflow then we are going to discuss about tensorflow with python and later on we are going to discuss about tensorflow with javascript and at the end we are going to discuss why should we use tensorflow so what is tensorflow basically tensorflow is an open source end-to-end machine learning platform it's a symbolic math toolkit that employs data flow and differentiable programming to handle a variety of tasks related to deep neural network training and inference it enables programmers to design machine learning applications utilizing a variety of tools libraries and open source resources google's tensorflow is currently the most well-known deep learning package on the planet machine learning is used by google in all of its product to improve search translation image captioning and recommendation for example google users can experience faster and more refined search experience with artificial intelligence if the user types a keyword in the search bar google provides recommendation about what could be the next word google wants to use machine learning to take advantage of their massive data sets to give their users the best experience when given a large amount of data deep learning began to surpass all other machine learning algorithms a few years ago google realized it it could improve its services by utilizing deep neural networks google is a popular search engine they created the tensorflow framework to allow researchers and developers to collaborate on ai models it can be used by a large number of individuals once it has been created and scaled it was first released in the late 2015 with the first stable version following in 2017. it's a free open source and thanks to apache for open source license without paying anything to google you can use it tweak it and redistribute the modified version for free data scientists and researchers programmers they all can collaborate and improve their efficiency by using the same toolbox tensorflow was built to scale because google has just more than data they also have the world's most powerful computer tensorflow is a machine learning and deep learning neural network research library created by the google brain team it was designed to run on variety of cpus and gpus and even on mobile operating systems now let's discuss the architecture for the tensorflow tensorflow allows you to create data flow graphs and structure to define how the data goes through a graph by taking inputs in a form of a multi-dimensional array called tensor it enables you to create a flowchart of operations that may be performed on these inputs which goes at one end and returns as output at the other tensorflow architecture is divided into following parts data preprocessing creating the model training the model and estimating the model tensorflow is named for the fact that it accepts input in the form of a multi-dimensional array commonly known as tensors you can create a flowchart of the process you want to run on the input which is called as graphs the input enters at the one end and passes through the system of various actions and finally exists at the other end as output tensorflow derives its name from the fact that tensor enters and flows through a set of operations that's why it finally exists now let's discuss about the tensorflow components tensorflow has following components tensor and graphs tensor gets a straight name from the primary framework tensor tensors are used in all of the computation of the tensorflow a tensor is basically an n-dimensional vector or a matrix that represents all kind of data a tensor's value all have the same data type and a known partially known shape the dimensionality of the matrix or array is determined by the geometry of data a tensor can be created from a raw data or outcome of a computation all operations in the tensorflow takes place within a graph the graph is a series of computation that occur at one and after other each operation is referred to as op node and they are all linked together let's discuss about graphs tensorflow employs a graph structure a graph collects and explains all series of computation performed during the training the graph has numerous advantages it was created to run on many cpus or gpus as well as mobile operating system the graph's portability enables the computations to be saved for immediate or future use graph can be saved or executed later all of these computations in the graph are accomplished by linking tensors a tensor consists of a node and an edge the node performs the mathematical action and generates the endpoint outputs the edge of the nodes explain their input output interactions now let's discuss about where can we run tensorflow tensorflow can be used anywhere tensorflow hardware and software requirements can be divided into two categories the mode is trained during the development phase and training is typically conducted on a desktop or a laptop computer tensorflow can be executed on a variety of platforms when the training phase is complete it is possible to run it on cloud as a web service desktop running windows mac os linux or ios and android mobile devices you can train it on machines and then execute it on a separate machine once a model is trained the model can be trained and used on gpus as well as cpus gpus were initially designed for video games in late 2010 strand for researchers found that gpu was also very good at matrix operations and algebra so that it makes them very fast for doing these kind of calculations deep learning relies on a lot of matrix multiplication tensorflow is very fast at computing the matrix multiplication because it is written in c plus plus although it is implemented in c plus tensorflow can be accessed and controlled by other languages mainly python finally a significant feature of tensorflow is a tensorboard the tensorboard enables to monitor graphically and visually what tensorflow is doing now let's discuss about few of the algorithms of tensorflow first we have linear regression then there is another algorithm named as linear classifier basically it's a classification algorithm then we have dnn classifier and at the same little bit it is modified into dnn linear combined classifier then we have boosted trees regressor algorithm and at the end we have boosted trees classifier now let's discuss about using tensorflow with python tensorflow makes all of these available to programmers via python language python is simple to learn and use and it provides convenient ways to define how high level abstractions might be combined tensorflow is supported on python versions 3.7 through 3.110 while it may function on older python versions but it is also not guaranteed tensorflow nodes and tensors are python objects and tensorflow applications are python applications however python does not do true maths operation the transformation libraries made accessible by tensorflow are developed in high performance c plus plus binaries python simply routes communication between the components providing high level programming abstractions to connect them the kira's library for example it is used for high level operations in tensorflow such as building nodes and layers connecting them the kira's api appears to be straightforward a basic model with three layers may be defined in less than 10 lines of code and training code is only a few lines longer however if you wish to lift the hood and undertake fine grained of work such as creating your own code you can also do that now let's discuss about tensorflow with javascript python is the most widely used programming language for tensorflow and machine learning in general however javascript is now a first class language for tensorflow and one of its major advantages is that it can run anywhere as there is a web browser tensorflow.js is a javascript tensorflow library that employs webgl api to accelerate computations using whatever gpus are present in the system it's also feasible to use web assembly back end for execution which is faster than usual javascript back-end if you are simply using a cpu through gpu are preferable wherever it is possible pre-built model allows you to get started quickly with the modest tasks to get a sense of how things function now the question comes up why should you use tensorflow tensorflow has many advantages abstraction is the single most important feature of tensorflow which gives for machine learning development instead of getting bogged down in specifics of implementing algorithms or finding out how to connect the output of one function to the input of another the developer may concentrate on the overall application of logic tensorflow provides extra benefits to the developer that needs to debug and get insights into tensorflow apps instead of generating the entire graph as a single opaque object and evaluating it all at once each graph actions can be evaluated and updated sequentially and openly this so called quick execution mode which was an option in the previous version of tensorflow is now the default through an interactive web-based dashboard the tensorflow visualization package allows you to monitor and profile the behavior of graphs tensorflow.dev hosted by google is a service that allows you to host and share tensorflow based machine learning experiments it's free to use with storage up to 100 mb scalers up to 1gb of tensor data and also 1gb of binary object data it is worth noting that the data hosted in tensorboard.dev is public so please do not use it for sensitive projects tensorflow also benefits greatly from a support of an a-list commercial entity like google google has fostered the projects of rapid development and offered numerous major solutions that make tensorflow easier to deploy and utilize one example is a previously stated tpu silicon for increased performance in the google's cloud thank you that was all from our site hope you would have enjoyed today's session and got some insights about what is tensorflow just a quick info guys intellipaat provides business intelligence masters program in partnership with microsoft the course link of which is given in the description below

Channel: Intellipaat

Views: 5,299

Keywords: tensorflow, tensorflow tutorial, what is tensorflow, introduction to tensorflow, tensorflow tutorial for beginners, tensorflow explained, learn tensorflow, how to use tensorflow, intro to tensorflow, learn tensorflow basics, tensorflow full tutorial, learn tensorflow for beginners, tensorflow for beginners, tensorflow extended, tensorflow and deep learning, why use tensorflow, what’s new in tensorflow, tensorflow training, tensorflow basics, tensorflow example, intellipaat

Id: Qsldxxo-7nI

Length: 11min 55sec (715 seconds)

Published: Tue Jun 21 2022