Machine Learning 101

Sometimes a topic like machine learning or AI (Artificial Intelligence) can be troublesome to understand without a STEM (Science, Technology, Engineering, and Mathematics) background, but today I am going to give a brief introduction to this topic and create the foundations in order to grasp the fundamentals of machine learning, let’s begin.

What is machine learning

In simple words, machine learning is a data technique that teaches a computer to learn from experience as we humans do. It sounds simpler than it actually is but this is the fundamental concept to understand. The computers are being trained to receive some samples data and use some analytics to learn from that and start to make their own decisions. It should not be confused with the term AI (Artificial Intelligence), although they are similar AI involves machines that do tasks that require human intelligence, like planning, understanding language, recognizing objects and sounds and learning.

AI has two main categories, general and narrow. General AI tries to achieve to have all the characteristics of human intelligence, and narrow AI achieves just one characteristic and does it extremely well, like recognizing images, but nothing else.

In this case, it is valid to say that machine learning is a way of achieving AI. now that is more clear the difference between AI and machine learning is easy to understand when someone is talking about these subjects.

The core thing of what machine learning does is to find patters in data. Then it uses those patterns to, in some way, predict the future, so it would be possible to detect when a new credit card transaction is likely to be fraudulent or what products to buy next.

To do this kind of learning possible and accurately it has to be enough data to find patterns, as we live now in a society that produces so much data it is possible to get better models for finding patterns and that is one reason machine learning is a hot topic today.

Why machine learning matter?

Right now machine learning is important for business leaders that want solutions to their business problems, also it is important for software developers for building better applications, and finally, it is important for data scientist, those are the ones who know about statistics, problem domain and how to write code.

With the rise of big data (larger and more complex data sets) machine learning plays a big part and helps to solve some problems is areas like computational finance, image processing, computational biology, energy production, manufacturing, and natural language processing.

How is the machine learning process?

The machine learning process is iterative, which means that things are repeated over and over in many ways, also it is not easy because sometimes it is difficult to find meaningful patters in large amounts of complex data, that is why data scientists are so important to machine learning projects.

The most important part of the process is to choose what question to ask, if the question to ask is wrong, the answer obtained is not going to be useful for the project. It is always important to question if the data gathered is the right data to answer the question.

In summary, this is the machine learning process:

1. Choose the data to work with.

2. Prepared data (Handles duplicates, missing data, and extra stuff).

3. Apply learning algorithms to the data.

4. Get a model result.

5. Iterate.

6. Get a good enough model to deploy.

7. Repeat the entire process over and over again.

A closer look at the machine learning process

Now is time to go a little bit deeper and explain how is the process of training data. The objective is to have a better understanding of the used terminology like supervised and unsupervised learning, classify machine learning problems and algorithms, how to train a model and test it, and so on.

Training data

Basically training data just means the prepared data that is used to create a model.

Supervised learning

Trains algorithms based on example input and output data that is labeled by humans, it means that the value that wants to predict is actually in the training data.

The purpose of supervised learning is to be able to learn by comparing its actual output with the taught outputs to find errors, in that way it uses those patterns to predict label values on additional unlabeled data.

A common use case of supervised learning is to use historical data to predict future events.

Unsupervised learning

Trains algorithms with no labeled data and finds structure within its input data, it means that the value that wants to predict is not in the training data.

The objective of unsupervised learning is to discover hidden patterns within a data set and is commonly used for transactional data, so for example when it is difficult for a human to make sense of the gathered dataset an unsupervised learning algorithm may determine some interesting patterns.

Regression, classification, and clustering

There are more categories but these are used a lot. Regression is used to examine the relationship between one dependent and one independent variable, enables prediction capabilities.

Classification is commonly used with supervised learning, where the data is needed to group it into classes, at least two or more and when it comes new data it has to determine which class that data belong to.

And clustering objective is to find clusters in the data with the use of unsupervised learning because there is no labeled data.

Decision tree learning, deep learning, and k-nearest neighbor

The decision trees learning, which is sometimes grouped together into decision forests, has their own neural networks, which imitate to some extend how the humans brain works. The goal of decision tree learning is to create a model that will predict the value of a target based on input variables.

One particular way of using neural networks is known as deep learning, this architecture is inspired by biological neural networks and consists of multiple layers in an artificial neural network made up of hardware and GPUs. Thanks to deep learning, computer vision and speech recognition have got significance advances and also potentially in the artificial intelligence space.

The k-nearest neighbor algorithm is a pattern recognition model that can be used for regression as well as classification, the input consists of the k closest training examples within a space.

Programming languages used in machine learning

There are some programming languages that can be used for machine learning like Python, Java, R, or C++, all of them have their own advantages and disadvantages. It is relevant at least to know one of them if the objective is to get involved in machine learning and want to create a career in this discipline.

Machine learning applications

There are many uses for machine learning, some good applications are Google Adwords bidding, product personalization, finding duplicate customer records in a database, personalize redemption recommendations in loyalty schemes, recommend the right product to the right person at the right time, accurately determine which of the marketing activities are having the biggest effect on sales, understand which marketing activities are most likely to move each individual customer closer to purchase, build and deploy machine learning algorithms that can detect anomalous behavior anywhere along with the blockchain, fraud detection, cybersecurity, build, deploy and refresh models to predict incoming threads in real-time, disease propensity, finding new oil and gas sources, credit card fraudulent transactions, and many more.

As you can see there are many applications of machine learning and is growing rapidly, this is a good time for learning and start to get more knowledge about it.

Ethical issues

The problem with machine learning is that the data can be biased, this has become a big problem because once you realize that the model could be biased it is hard to see why it is doing what it is doing, for example, the algorithmic bias in criminal justice systems, in this situation once it start using data about crimes committed by African Americans, the prediction model would start to be biased against black communities, in a big-picture if minorities are underrepresented in the data samples the models would always have some kind of bias.

Conclusion

I hope this overview of what machine learning is can help to get the fundamentals and a broader picture of this topic and open the path for deeper knowledge and begin to speak the same language as someone with a strong background in science, technology, and engineering.

Sources

• An introduction to Machine Learning
• The difference between artificial intelligence, Machine learning, and Deep learning

Regresar

Enjoy this note? Feel free to share!