Machine learning is a sub-domain of artificial intelligence. This refers to a set of mathematical and data transformation techniques that allow a software to learn from historical data in order to automatically adjust to perfect the performance of a specific task.
The word learning in machine learning is a metaphor for how human beings (or any living being with an intelligence) can gradually improve their performance through observations and lessons learned from past experiences.
Machine learning has several advantages that justify its growing popularity.
The availability of data and the increase in computing capacity make it possible to build «intelligent» software capable of solving increasingly complex and varied problems.
Machine learning allows computers to continuously learn and improve their learning and, in some cases, to surpass human capabilities.
Example: identification of spam
To illustrate this concept simply, let’s take the example of one of the first popular uses of machine learning: the identification of spam (spam). Let’s say we wanted to build a computer program that could recognize spam among the hundreds of emails we receive every day.
The first reflex would be to establish the recognition characteristics of spam, such as a high number of hyperlinks, an abusive use of commercial language, etc., which we will call feature extraction.
The next step would be to manually introduce a set of rules (e.g. if the number of hyperlinks is greater than seven, report the email). In reality, this approach can be very limited.
Since the structure and content of spam can be extremely varied, the number of features will be very high.
The manual introduction of all the rules necessary to identify spam would then be impossible.
It would be more appropriate to use a machine learning algorithm to which we would introduce a data set consisting of emails allowing us to clearly distinguish spam, which we will call training data (training set) and train the algorithm to automatically identify the characteristics associated with spam.
We will then test our algorithm’s ability to recognize spam using a test set (testing set) that consists of a new set of emails (different from the emails used during the training phase).
This will test the actual ability of our software to identify spam.
In this case of use, the choice of machine learning offers another advantage: the continuous improvement of the detection performance.
New emails can indeed be used to improve the knowledge base of our computer program and thus fuel it with new features to consider in the future.
How to build machine learning algorithms in Python:
Last year, an AI known as AlphaGo beat the world’s number one go player. This was made possible by a machine learning model that learns to understand trends and data models and, in this case, Go moves. Machine learning has its uses outside of fun and games; it can be applied to any domain and you can learn to apply your own machine learning models with this Python Machine Learning ebook, currently on sale for $10.
This resource includes 454 pages of content, which will teach you how to create Python machine learning algorithms to generate useful information. You will learn how to apply different machine learning approaches to different data analysis questions, as well as how to create code to optimize the strength of machine learning algorithms. You may use your new skills to integrate machine learning models into web applications.
The Python Machine Learning guide costs $32.99, but you can sell it today for $10, saving almost 70% on the usual. price.
Limitations of Machine Learning:
The algorithm can only learn from available and known learning data. Thus, as in our Data Set, silicon values range from 5% to 8.5%, so the model will not be able to predict the behavior of hypersilicon alloys (Si to 17-25%). In the same way, unknown aluminium alloys (for example doped in scandium, nickel or zirconium) not present in the learning base will be outside the scope of the model. Conclusions
We have described, based on an aluminum Data Set available at the CTIF, how to make Machine Learning under Python with the relatively powerful standard libraries of Scikit-Learn. Beyond the use of libraries, the preparation (preprocessing) of data is very important with different steps (cleaning, encoding, normalization). We deliberately omitted to mention neural networks, as this type of ML algorithm will be discussed in another article.
keywords: machine learning, machine learning is, python machine learning,machine learning modeling, andrew ng machine learning , ai learning , aws machine learning, supervised learning ,unsupervised learning, ai ml, deep learning ai, tensorflow, data analytics, master's in data science, online master's data science, data analytics degrees, data science degrees, certified data scientist, master's in data analytics online , ms in data science, datascience berkeley ,uc berkeley data science, data science for managers, data science for beginners, certified data scientist, data science for all, big data analyst, r for data science, pandas, keras,tensorflowjs,hands on machine learning.
crossorigin="anonymous">