Learning Machine Learning 0.1


Machine Learning seen by a beginner: the basic key concepts

First question: what is Machine Learning?

In an easy way, I would say that Machine Learning is a mix between math, statistics and a few of AI that when facing a question or a problem finds patterns and provides a prediction. This prediction could be a value or a set of values and they come from the analysis of a bunch of related data that we must collect previously and then is processed by ML.

What is machine learning?

How to predict?

As I said before Machine Learning uses essentially math and statistics to predict values. There are lots of algorithms and techniques that we can explore and use for it. Basically our job as ML engineers is to find which one or which that fit most appropriately with your scope project.

The chosen algorithms should be part of the model as well as our dataset, the set of context samples. The model also contains the flow that our data should follow in the learning process and the output should be our expected prediction.

Classifying, Regressing or Clustering?

There are lots of ways to get the learning from our dataset depending the kind of data and also the problem we have. If our dataset consists in a bundle of observations and we have a target to predict, we are facing a supervised learning and we can handle it with classification (target is a label or a class) or regression (target is a continuous variable).

Is the cat sleepy or active?
Classification: Is the cat sleepy or active?
How long is the cat sleepy?
Regression: How long is the cat sleepy?

On the other hand, if we don’t have a target we’ll gain unsupervised learning by grouping or clustering data by their similarities.

Clustering: How can we group them?
Clustering: How can we group them?

Which data?

For each problem, we need to get a bunch of data so we can learn with that. This data should be a set of observations or samples composed by one or more attributes related to our problem. After applying that all math and statistic we’ll need a different set of data so we can test the learning achieved. In order that we use to separate our dataset in a training set (the big one) and a testing set (the small one).

How to predict better?

How to predict… better?

I believe the secret of Machine Learning is to find the most appropriate algorithms for our problem. This accuracy comes with persistency and experience so as we become familiar with the algorithms and refining our models, the predictions will become increasingly credible and closer to the reality.

This is my first step in Machine Learning and I’m aware that there still is a long path in my journey.

I’m a very very very beginner in the Machine Learning world and in the middle of my studies after the first impact with this subject, I decided to put in paper (digital paper?) what I think could be the basic key concepts of ML. If I am able to explain these concepts, it means that I understand them, right? I’ve been reading some articles (I’m following Get Started In Machine Learning in 5 Steps) and learning from tutorials I’ve come across but the goal of this post is to get some feedback and more knowledge from your experience. I would be glad if you could take a few minutes of your time to help me out clarifying my ideas and to improve this share with some important concepts that I’ve missed.

Some Sources

Machine Learning — What it is and why it should interest you!
Machine learning is the concept of using algorithms to identify patterns and / or make predictions based on an input…towardsdatascience.com

Get Started In Machine Learning in 5 Steps
Machine learning is the biggest enabling technology since electricity. Every day there is a new advance in the field.

An introduction to machine learning with scikit-learn - scikit-learn 0.19.1 documentation
unsupervised learning, in which the training data consists of a set of input vectors x without any corresponding target.