nanosense-blog
en English

What is Machine Learning? Difference between Machine Learning and Artificial Intelligence

Share on facebook
Share on twitter
Share on linkedin

How do we recognize objects when they look the same in plain view? By touching them? Texture-wise, maybe no difference can be found. By tasting them? Not all objects can be felt. What about smelling them? They may have different smells, but this only happens when they possess sensitive senses. So, how can we recognize multiple objects with high precision? Surely, we can use machine learning!

Then, how is machine learning able to recognize certain objects? How does it work? And what exactly is machine learning itself?

 

What is Machine Learning?

 

Machine learning (ML) is an automated method of data analysis in the creation of analytical models. ML is a branch of artificial intelligence (AI) and computer science that focuses on using data and algorithms, identifying patterns, and making decisions that mimics human ways, and gradually improve their accuracy.

According to IBM, ML is an important component in the evolving field of data science. ML algorithms are trained to make decisions about classifications or predictions and uncover key insights in data mining projects using statistical methods. These insights further drive decision-making in the application, which ideally affects the growth metrics of the main data.

In a nutshell, ML can be interpreted as:

  • A machine that can learn from the data itself.
  • A data analysis method that automates the development of analytical models, using algorithms that learn from data.
  • A practice of training computers through algorithms to recognize patterns and infer predictions that mimic a human-like ability to learn from “experience.”

 

Read also:

 

 

Position of Machine Learning in Artificial Intelligence

 

Machine learning (ML) is a branch of science from artificial intelligence (AI). However, for their application, they are very different, which can be seen in terms of:

Applicable Purpose. AI is employed to increase the chances of success and solve complex problems with natural intelligence simulations, while ML aims to improve efficiency without being success-oriented and needing to learn from data to improve machine or system performance.

Algorithms. AI mimics human capabilities in terms of response and behavior for systems. Meanwhile, ML, is able to create its own algorithm for the learning process.

 

Since Machine learning (ML) and deep learning (DL) tend to be used interchangeably, it should be noticed that both of them are different in context and their usability. Machine learning (ML), deep learning (DL), and neural networks (NNs) are all subfields of AI. DL is actually a subfield of machine learning, while NNs is a subfield of deep learning.

Deep Learning. DL combines advances in computing power and types of neural networks to study complex patterns in large amounts of data. DL techniques, for now, are tools to identify objects in images and words in sound.

Neural Networks. NNs also known as artificial neural network (ANNs), consists of a node layer, which contains an input layer, one or more hidden layers, and an output layer. Each node, or artificial neuron, is connected to the others and has an associated weight and threshold. If the output of any node is above the specified threshold value, that node is activated, sending data to the next layer of the network. Otherwise, there is no data passed to the next layer of the network. A neural network consisting of more than three layers—which includes input and output—can be considered a deep learning algorithm or a deep neural network. A neural network that only has two or three layers is called basic neural network.

source: hackernoon.com

 

Why is Machine Learning Important?

 

Machine learning (ML) and data mining (DM) (a component of machine learning) are critical tools in the process of gathering insights from the large data sets held by companies and researchers today. This means that it is possible to quickly and automatically generate models that can analyze larger and more complex data as well as provide faster and more accurate results – even at very large scales. By building the right model, an agency or organization has a better chance of identifying more profitable opportunities – or avoiding unknown risks.

 

How Does Machine Learning Work?

 

Basically, the working principle of machine learning (ML) is similar to that of artificial intelligence (AI), which includes learning, reasoning, and self-correction. To understand how ML works, there are some commonly used terms that need to be known, i.e., decision process, error function, and model optimization process.

Decision. In general, ML algorithms are used to make predictions or classifications. Data input needs to be determined at the beginning, whether the data can be labeled or not, the algorithm will generate a pattern estimation in the data.

Error Function. An error function is employed to evaluate the prediction of the model or case classification. If there are known examples, then this serves to detect false cases or can make comparison to assess the results against the accuracy of the model.

Model Optimization. If the model can better fit the data points in the training set, then the weights are adjusted to reduce the difference between the known sample and the estimated model. The algorithm will repeat this evaluation and optimization process, updating the weights independently until the accuracy of the threshold is found.

 

source: Pixabay.com

 

Methods in Machine Learning

 

Many machine learning models are defined by the presence or absence of human influence on the raw data. Generally, ML has three basic learning techniques, i.e., supervised, unsupervised, and semi-supervised.

Supervised Learning. Supervised machine learning is the use of labeled data sets to train algorithms for classifying data or predicting results accurately. For example, a device may have data points labeled F (failed) or R (runs). When input data is entered into the model, it adjusts its weights until the model is fitted correctly, by comparing the actual output with the correct output to find errors. This occurs as part of the cross-validation process to ensure that the model avoids overfitting or underfitting. Supervised machine learning is typically used in applications where historical data predicts possible future events, such as classifying spam in a separate folder from your inbox. Some of the methods used in supervised machine learning include neural networks, Bayesian naves, linear regression, logistic regression, random forests, support vector machines (SVM), and many more.

Unsupervised Learning. Unsupervised machine learning is the use of machine learning algorithms to analyze and group unlabeled data sets (as opposed to supervised machine learning). These algorithms find hidden patterns or clusters without being told the “right answer”, so the algorithm has to figure out what is being displayed on its own. The goal is to explore the data and find some structure in it. Unsupervised machine learning works well with transactional data, due to its ability to spot similarities and differences in information, making it an ideal solution for data analysis. It is also used to reduce the number of features in the model through a dimensionality reduction process; Principal component analysis (PCA) and single value decomposition (SVD) are two common approaches to this type of learning. Other algorithms used in unsupervised machine learning include self-organizing maps, nearest-neighbor mapping, k-means clustering, singular value decomposition, neural networks, probabilistic clustering methods, and more.

Semi-supervised learning. Semisupervised machine learning is utilized for the same applications as supervised machine learning. However, it uses smaller amounts of labeled data and larger amounts of unlabeled data for algorithm training. This type of learning can be used with methods such as classification, regression, and prediction. Semisupervised machine learning can solve the problem of not having enough labeled data (or unable to provide sufficient label data) to train supervised learning algorithms.

 

What Are the Applications of Machine Learning?

 

The applications of machine learning (ML) can help humans in many fields. Even today, you can easily find the application of ML in everyday life. For example, when you use the face unlock feature to unlock your smartphone, or when you browse the internet or social media, you will often be presented with advertisements (ads). The appearing ads are resulting from ML processing which will provide advertisements according to your personality.

On the other hand, most industries working with big data have recognized the value of ML technology. By gathering insights from the data – often in real-time – an institution or organization can work more efficiently, even gaining an edge over competitors.

 

Source: Pixabay.com

 

Challenges of Machine Learning in the Future

Machine learning (ML) algorithms are used worldwide in almost every major sector, i.e. business, government, finance, agriculture, transportation, cybersecurity, and marketing. Its rapid adoption across different industries is a testament to the value created by ML. 

Again, the challenges faced by ML are similar to those that must be faced by artificial intelligence (AI), such as:

1. Data bias

The ML algorithm is based on data, and it is undeniable that data certainly have a data bias. If humans compete with each other trying to develop algorithms to correct biases, then wouldn’t that disrupt the current world system? Or is it enough to just rely on the data regardless of bias?

2. Data security or privacy protection

If all the data we need to build an ML system are available, then who can access them? What can the parties do with certain data? And when it comes to personal privacy data, is privacy protected enough?

3. Responsibility

If an ML algorithm makes a decision, then who will be responsible for the false decision? For example in the cases of self-driving cars (autopilot), medical diagnosis, and facial recognition, who will be responsible if something goes wrong or there is an accident?

 

These three challenges cannot be solved by technological answers themselves. Instead, they also need to be answered from political and legal views as well.

 

Source:

https://www.ibm.com/cloud/learn/machine-learning#toc-machine-le-K7VszOk6

https://www.sas.com/en_id/insights/analytics/machine-learning.html

https://ischoolonline.berkeley.edu/blog/what-is-machine-learning/

Share on facebook
Share on twitter
Share on linkedin

Newsletter

Want to get more inspiration from Nanosense?
Come on, sign up for our newsletter!

Leave a Reply

Your email address will not be published. Required fields are marked *

Editor's Pick

You can change the display language through the
features that available in the upper right corner.

Anda dapat mengubah bahasa tampilan melalui
fitur yang tersedia di pojok kanan atas.