What is Machine Learning?

Machine Learning San Francisco

If you’ve been considering taking data classes, then you’ve probably heard of artificial intelligence. What you might not know is that the term was coined in 1955 to describe new types of computer science. Now, machine learning has gained popularity, and many people consider it to be almost as reliable as artificial intelligence. Because these two things are so closely linked, it’s vital that you understand Python machine learning before moving on to artificial intelligence. Python is often used for machine learning because it is an approachable and familiar language.

What Role Does Python Have in Machine Learning?

Most data classes will teach you that machine learning is more connected to probability than logic or reason. Consider these examples: When you see someone’s face, your brain is able to estimate the likelihood of your having seen that face before; when you play a game, your brain can determine which strategy will be likely to help you win. While these tasks each involve intelligence, they also involve probability.

Machine learning uses intelligence and probability in the same way your brain does. If a computer has been provided enough data, then it can easily estimate the probability of a given situation. This is how computers are able to recognize photos of people on Facebook and how smart speakers understand commands given to them.

So, how does Python machine learning estimate probability? First, it breaks problems down into small pieces of information that can then be linked together to create applications and learn to perform many types of tasks. Let’s look at an example to better understand how machine learning works.

Imagine that you want to create a machine learning algorithm to see if an image contains a human face. You could use a neural network, which has many layers of neurons. The neurons in the first layer may learn to seek out a shape, such as a line or a curve, while another layer may see if the lines or curves detected by the first layer make up advanced shapes, which include circles or corners. Each layer will then look for progressively more advanced shapes, including human features, like eyes or noses. The final layer’s neurons communicate with the machine learning algorithm. The algorithm is then able to estimate the likelihood that the image contains a face.

One of the most advantageous parts of Python machine learning is that the computer’s algorithm can make these estimations on its own. All the user must do is to add images into the algorithm and make a few parameters which might include the number of layers and neurons. The machine learning algorithm can then make a guess based on the information found by the layers. This process is known as machine learning because with every guess, the algorithm learns to make a better guess the next time.

Applications of Machine Learning

If you’re interested in learning about data science, it’s good to understand more applications of machine learning. This article will take a look at speech and image recognition, robotics and reasoning.

Speech Recognition 

You have most likely used speech recognition at some point in your life, whether you have a smart speaker or have used the “Hey, Google” or “Hey, Siri” feature on your smartphone. The speech recognition process is also referred to as natural language processing (NLP). Because machine learning has grown immensely in the past decades, companies such as Amazon, Apple and Google have all worked with applications for speech recognition.

Image Recognition

We’ve already started to look at how machine learning works for image recognition. For instance, social media sites that automatically tag photos of you and your friends use image recognition to do so. However, machine learning for image recognition is not without its faults. Because a computer image algorithm is a statistical relationship between the pixels, there’s always the potential for error. If you put stickers on a stop sign, then deep learning machines may not recognize the sign for what it is. The algorithm’s knowledge of what stop signs look like is based on past images that have been fed to it. When the appearance of the image is altered in some way, the machine learning algorithm may have trouble matching it to past knowledge and recognizing what is in the image.


Another way that you can use Python machine learning is through robotics. Recognizing the identity of an object or person is one thing, but manipulating the input takes machine learning to a whole new level. However, it’s complicated because the field of robotics is still being developed and manipulating an object involves many steps. 

At first, professionals tried to overcome these complexities by creating programs that provided step-by-step instructions to the robotic arms. However, there are too many variations in the way that an object can be picked up, and writing the code for each variation would require too much time. 

Instead, the computer must recognize which object to move and then estimate how much force is required to move it. Then the computer must estimate where the item must be grasped and how much force should be applied to grasp it. Each of these tasks involves uncertainty, which is multiplied across the entire task. 

Most data classes will explain the difficulties of manipulating objects with robotics. This is why it’s crucial for a machine to be able to learn how to interact with the world on its own. For now, machine learning algorithms that will give a robot the same qualities and abilities as humans are still in the future. 


Another aspect of machine learning is reasoning, which is currently still in development. The best strategy to build artificial intelligence that can reason involves reinforcement learning, however, reinforcement learning requires human intervention so the outcome can be defined. Reinforcement learning is not useful when it comes to areas where the outcome is not clear. For example, it could not be used for questions such as, “What should our company’s growth strategy be?” These limitations and others mean that machine learning still has some progress to make before it can take over intelligence-related tasks, such as answering difficult questions. 

The Future of Machine Learning 

Your data classes may explain that machine learning algorithms can get close to the intelligence of humans when it comes to many tasks, including speech recognition. While machine learning is successful, it will likely not take the place of human intelligence anytime soon. One of the key questions about machine learning is how much human intelligence can be approximated with statistics.

However, even if no more discoveries in machine learning were to arise, the algorithms that exist today still have a multitude of uses. Because of the advances in machine learning, such as robotics, vision and speech, the world can still be reshaped with the help of this technology.

A century ago, the world did not have computers. Imagine the advances in machine learning that could take place in the coming century and how they can make the world a better place. Whether you are planning on taking data classes in San Francisco or are studying on your own, you can benefit from knowing about Python machine learning — the future of technology.

*Please note, these articles are for educational purposes and the topics covered may not be representative of the curriculum covered in our boot camp. Explore our curriculum to see what you’ll learn in our program.

Get Program Info


Ready to learn more about Berkeley Data Analytics Boot Camp in San Francisco? Contact an admissions advisor at (510) 306-1218.