The concept of machine learning

 The concept of machine learning

The concept of machine learning

How does machine learning work?


How do we learn, what does "learning" mean, and what does it mean for a machine?

 

Computer scientists, mathematicians, neurologists, educationalists, philosophers, and artists are all fascinated by the issue of learning.

 

One description offered by Fabien Benureau (2015) is "Learning is a modification of behavior on the basis of experience," which holds true for computer programs just as much as it does for robots, animals, and people.


How does machine learning work?


 When a computer program has the ability to learn without this alteration being explicitly written, which is what this article is interested in, we speak about machine learning.


As a result, we may compare a traditional program—which utilizes a procedure and the data it receives as input to produce answers as an output—with a machine-learning program, which uses the data and the answers to create the procedure that derives the latter from the former.


Example


Let's say a business wishes to determine from a customer's invoices the total amount they spent. Applying a traditional method, such as basic addition, is all that is necessary; no learning algorithm is needed.


Now imagine that we want to use these bills to ascertain which things the client is most likely to purchase within the next month. We obviously don't have all the information we need to perform this, despite the fact that this is probably linked. 


However, if we have access to a large number of people's purchasing histories, we can apply a machine learning algorithm to extract a prediction model from them and answer our question.


What can machine learning be used for?


Machine learning can be used to solve a variety of problems, including those that we don't know how to solve (like the shopping prediction example above), those we do know how to solve but can't formalize in algorithmic terms (like image recognition or natural language comprehension), and those we know how to solve but with methods that are way too resource-hungry (like the prediction of interactions between objects, for example).


In situations where expertise is scarce or underdeveloped but data is reasonably abundant, machine learning is applied.


The models produced by learning algorithms can highlight the relative relevance of different pieces of information or how they interact to solve a specific problem. In this way, machine learning can also aid human learning.


Understanding the model can help us identify the traits of past purchases that are indicative of future ones, like in the case of purchase prediction. This use of machine learning is commonly employed in the study of how genes contribute to the growth of various tumor types.


Which brain areas can be used to predict behavior? What features of a molecule make it an effective medication for a certain indication? How can a specific astronomical object be identified from a telescope image?


Machine learning components :



machines learning components


The two main pillars of machine learning are the data and the learning algorithm. The data serves as the examples from which the algorithm will learn, and the algorithm serves as the process we use to process the data to create a model. Training is the process of applying a learning algorithm to a set of data.


These two pillars are both essential. The garbage in, garbage out principle states that a learning algorithm given poor-quality data will be able to do nothing with it other than produce poor-quality predictions. 


On the other hand, no learning algorithm will be able to create a good model from irrelevant data. However, a model that was learned using an inappropriate technique on pertinent data cannot be of high quality.


Remember that a machine learner's or data scientist's job involves engineering work to prepare the data in order to remove outliers, handle missing data, select an appropriate representation, and other tasks.


Attention


It's vital to distinguish between a machine learning algorithm and a learned model even though it's common to refer to the two as the same thing. The former uses data to build the latter, which can then be implemented like a regular program.


As a result, a phenomenon can be modeled using a learning algorithm using instances. We must define and optimize an objective in order to accomplish this. This can involve, for instance, reducing the amount of mistakes the model makes when learning instances.


Example


Here are a few illustrations of how optimization challenges might be used to reframe machine learning issues.


- By optimizing the proximity of customers assigned to the same type, an online merchant may attempt to model representative client kinds based on historical transactions;

- to reduce the number of accidents, a car corporation would seek to model a vehicle's trajectory in its environment using video records of cars;

- To maximize the compatibility of their model with current information, genetic researchers may aim to model the effect of a mutation on a disease using patient data;

- A bank could wish to simulate hazardous conduct in order to increase the likelihood that non-solvency will be discovered.


Therefore, machine learning depends on computer science for the encoding of data and the effective use of optimization algorithms, as well as mathematics, particularly statistics, for the development of models and their inference from data. The utilization of distributed computing and database designs is becoming more and more necessary due to the sheer amount of data that is available.


The set of approaches used to create computers capable of exhibiting behavior that may be characterized as intelligent is what is referred to as machine learning, which can be thought of as a subset of artificial intelligence. Indeed, it is difficult to consider a system clever if it is incapable of learning. A system that is intended to adapt to a changing environment must have the capacity to learn from experience and profit from it.


Solutions for open-source cloud computing




Comments