Hello, people I hope you all are having an amazing time learning, I am back with this post that will give a complete introduction to different aspects of machine learning which is a new trend in the world of cutting edge technologies. Before diving into details let's take a look at what is machine learning and how it works. # What is machine learning? >"Machine learning is an application of AI that equips the systems with the ability to automatically learn and improve from experience without being explicitly programmed." Now let me explain it to you, as you know we can program computers to do complex calculations and computations and provide us with desired results. Artificial intelligence is a quest to induce humans like intelligence in machines so we no longer have to program each detail. Machine learning is also an application of AI that aims to program computers in such a way that they will learn and improve by themselves, we only need to provide them with enormous amounts of data that they would use to improve without human intervention, assistance, or any explicit programming. The learning is based upon observations or data, i.e. direct experience, or instruction, to refine the output as much as possible. A simple example of machine learning is your google assistant installed on your phone, the Natural Language Programming (NLP) of the assistant uses the technique of machine learning to remove ambiguity from user's query and analyzing each word according to the context provided (semantic analysis) rather than on its literal meaning. https://universityofreston.com/wp-content/uploads/2020/09/CX2-1.jpg machine-learning ## Methods of machine learning The algorithms used for ML are mainly categorized as supervised learning or unsupervised learning. * ### Supervised Learning These type of algorithms works by applying what is learned in past to the new data provide to predict future events. It does so by comparing its prediction to correct intended outputs and modifies its model according to the accuracy. It begins with analyzing the known dataset including labeled examples and after enough iterations, it generates an inferred function to predict the outputs and after sufficient training, it finally becomes able to predict output for new unknown inputs. https://cdn-images-1.medium.com/max/1600/1*Iz7bCLrPTImnBDOOEyE3LA.png supervised machine-learning * ### Unsupervised Learning The method of unsupervised learning is applied when the data used to train the algorithms is neither labeled nor classified. It analyzes unlabeled data and provides us a way to infer a function to describe an ambiguous structure. The algorithm does not generate accurate outputs instead it works on the data to make inferences from provided datasets to extract hidden structures. https://cdn-images-1.medium.com/max/1600/1*C7meM0f0_oy8RPM9HDb5gg.png supervised machine-learning * ### Semi-supervised Learning The algorithms based on semi-supervised machine learning algorithms that's a middle way between supervised and unsupervised learning, since they use mixed (both labeled and unlabeled data) for training – majorly consisting of unlabeled data and a small share of labeled data. The learning accuracy of these algorithms are considerably improved by applying this method to systems. Generally, this learning technique is applied when the gathered labeled data needs skilled and relevant resources to train the algorithm. Otherwise, gathering unlabeled data doesn’t require futher resources. https://cdn-images-1.medium.com/max/1600/1*XkhJYY6-oRYGq6s0Txfj1g.png semi-supervised machine-learning * ### Reinforcement machine Learning This learning method allows algorithm to interacts with its surrounding by making actions and finding errors. Trial and error and delayed reward are few prominent characteristics of reinforcement machine learning. This technique allows the systems to find out the ideal behavior within a specified context for maximizing accuracy. This algorithm applies simple reward feedback to determine the best results; this is called reinforcement signal. https://cdn-images-1.medium.com/max/1600/1*5ugO5IDdynfFpgTa2vkw6Q.png reinforcement machine-learning ## Some uses of machine learning * Social media websites use machine learning algorithms to provide more personalized user experience. You might go somewhere, see someone out there, when you reach home you open your social media and suddenly that guy appears in people you may know category, this is not magic, this is algorithms in action. * Sales and advertisements companies also uses machine learning to increase their sales, you may search for something on google and suddenly you will start to see advertisements on every app about that particular product. * Brands and websites uses machine learning to train their virtual assistants to provide customer services and support. You may message some brands facebook page and receive replies in seconds, it is not because someone is always sitting there for you, its the chat bot helping you with your query. * CCTV cameras and surveillance also uses machine learning to monitor the activities at every nook of the area without any human intervention. * Automatic speech recognition (ASR) uses natural language processing (NLP) to process human speech into a written format. Many applications use speech recognition into their systems to take instructions by voice —e.g. Google Assistant, Bixby, Siri, etc. ## Conslusion Machine learning tries to imitate the way human learn new things by observing and trial and error method. Machine learning allows us to analyze enormous quantities of data and generates faster, more accurate results to discover opportunities and avoiding risks, though it requires time and resources to train the algorithm. Combination of ML with AI and cognitive technologies can make it highly effective in processing enormous quantity of data.