What is Machine Learning?

To explain Machine Learning, we should start by asking the question: What are the things, commodities, tasks, actions or tools that machines can do for humans? Machines can execute in a more efficient manner than humans, but in order to do so we humans must train machines for them to start learning. Machine learning is the General Term for when Computers learn from data. There are different ways (algorithms) in which can learn: the algorithms can be grouped into supervised and unsupervised learning.

‍

Let’s get into what supervised learning is:

‍

To understand what supervised learning is we need to take into consideration for the algorithm to learn, first it needs input data and secondly the desired output data.

Let’s go with an example to light our idea better:

Based on your past music choices, your music service platform recommends new songs. But how is this supervised learning?

Well, the model is training a classifier on pre-existing labels. Features like genre, tempo or any other characteristic a song can have can be attributed to a song which then can be considered a label. I.e. A Madonna song can be labeled by having the features of Pop Music, 80’s Music, down or fast tempo, etc. Based on those features the algorithm can predict other songs (labels) that share either the same or most of the features.

‍

Key Points on Supervised Learning for Machines:

1. The algorithm is fed by an input, then an associated output.
2. Repeat the previous step as many times as possible.
3. The algorithm picks up the pattern between input and output.
4. The algorithm is ready to be fed with a new input and it will be able to predict the outcome.

‍

Some popular applications for Supervised Learning would be:

1. Predictive Analytics i.e. The Stock Market Exchange Prices
2. Text Recognition
3. Spam Detection
4. Object Detection i.e. Face recognition on images.

‍

Now, let’s go into unsupervised learning:

‍

Unsupervised learning differs from supervised learning in that data is not labeled. Meaning algorithms are only fed with input data. 

Let’s go with an example to light our idea better:

A payment platform which regularly receives X number of payments, one day starts receiving a higher amount of payments, with all kinds of values. The owners of the platform need to know whether these payments are legit or they are fraud. In order for the algorithm to know accurately it has to look at the input data; i.e. Value of payment, time between transactions, IP of transactions, etc. All of this data is computed to draw patterns and seek for outliers. Since the new transactions were not labeled this accounts for Unsupervised Learning.

‍

Key Points on Unsupervised Learning for Machines:

‍

1. Algorithm is fed with an input example.
2. Repeat the previous step as many times as possible.
3. The algorithm picks up on patterns and starts clustering data.
4. A new input is fed on the algorithm, based on preexisting data it will cluster the new input where it belongs.

‍

Some popular applications for Unsupervised Learning would be:

‍

1. Data Visualization
2. Structure Discovery
3. Targeted Marketing
4. Customer Segmentation

‍

To answer our initial question: What are the things, commodities, tasks, actions or tools that machines can do for humans?

The applications are almost infinite, just as human behavior & activities (and beyond everything concerning us humans machine learning will be looking at all of life all together).

We at Remotasks work with different kinds of models. And if you are interested in building the future for humanity, Tech and Machine Learning, you can join our platform to help companies build the AI of the future.

JOIN REMOTASKS TO START TASKING AND EARNING MONEY BY HELPING BUILD THE AI OF THE FUTURE. CLICK HERE

‍