What Is Machine Learning | Deep Learning Vs Machine Learning Vs Artificial Intelligence

Definition of Machine Learning

Computer ability to learn from data is Machine Learning. Machines learn from traditional programming methods wherein data is used. Traditional methods like decision trees are based on statements or case statements. However nowadays, Machine Learning allows computers to learn from patterns that are available in images and data. Pixel values of the actual image enables the computer to identify and differentiate between various options and give the right answer. Earlier methods used labelled and unlabeled data. Now, Machine Learning uses a combination of data and images.

Example: An integer could easily be differentiated from a string based on data input in the computer in the required programming language and in the traditional programming method. However, to identify an apple or an orange it has become necessary to integrate data with picture images to provide the right answer.

Areas or Services where Machine Learning is used

In 2019, Machine Learning has permeated many areas and services. An extensive list is given below. It powers decisions in

• the field of finance, stocks and high frequency trades
• Advertising and targeted ads are also impacted by machine learning
• Computer vision in self driving cars
• Dealing with text to help in sentiment analysis. This determines if the context of the text is happy, sad or angry and helps to gauge the mood of the words
• Fraud detection
• Weather forecasting and prediction
• Medicine
• Robotics
• Video games
• Speech to text wherein audio is translated to text and vice versa, from text to audio
• Recommendation models wherein machine learning is used to recommend one thing over the other when they are closely paired or matched or with very minute variations
• Accurate and efficient translation in various languages to enable ease of communication between all people in the world and erase language barriers

Machine Learning Vs Deep Learning

Machine Learning uses many different types of algorithms. A concept known as neural network is also an algorithm. The name derives from the network of neurons in the human brain. This neural network scales out and helps to develop more complex models to handle larger and bigger problems.

Deep Learning is a subset of machine learning. Deep Learning tends to over complicate things. Sometimes, a simple Machine Learning algorithm like K-means clustering or simple logistic regression may be enough for your needs but you tend to resort to Deep Learning models to handle these problems. Deep Learning models are expensive computationally and require a lot of resources to maintain and manage.

They are extremely useful for many problems and specifically in the realm of computer vision and text problems. They are the best solution for these problems. However, it is best to know through trial and error method or by consulting people who may have worked on similar problems in the past or by researching online if Deep Learning models are needed to solve the problem.

Machine Learning Vs Artificial Intelligence

Artificial intelligence is a constantly changing and evolving term. Over the years, many changes have come in and the term has a broader purview today. The definition of this term is encompassing and Machine Learning is a subset of Artificial Intelligence.

Many tools which are known as Artificial Intelligence today are powered by Machine Learning algorithms. Machine Learning, while primarily being a subset of Artificial Intelligence, retains its characteristic uniqueness.

The nature and simplicity of some Machine Learning algorithms do not allow it to be referred to as Artificial Intelligence. Machine Learning remains a more contentious and well-defined term than Artificial Intelligence.

Machine Learning is a broad term. While it falls under the larger category of Artificial Intelligence, it has its own well-defined parameter. Deep Learning is a subset of Machine Learning and Artificial Intelligence. All these categories have a deep imprint on our daily lives and have a huge presence in all areas of the world economy and industry.

Nikunj Bajaj, the senior Machine Learning Engineer at Facebook, shares his knowledge and draws upon his own personal experiences in the field of Machine Learning to share some valuable insights into this fascinating and trending topic.

What is Machine Learning

Machine Learning is a technique to empower machines to learn. The name itself implies this. It is a specific task based on data. It begins with your having a goal in mind and then programming the machine to learn that specific task based on previous experiences which is provided to the machine as data.

      For example, if you want to teach the machine to recognise digits, the experience you provide to the machine will be to give it the letters printed on a paper. You can even provide it images of the letters and even label it to show that each letter corresponds to a specified digit. By doing this you empower the machine to learn the digits on its own so that when a new paper or image is given to the machine it can figure out what digit it is.

      The goal of Machine Learning is to ensure that the machine makes minimal error while being able to connect the mode accurately. That is the learning the machine is getting. Essentially, any technique that educates or empowers the machine to learn like humans would do, is Machine Learning.

Machine Learning Vs Deep Learning Vs Artificial Intelligence

• Deep Learning Vs Machine Learning
  
  • Deep Learning is a subset of Machine Learning. While Machine Learning is any technique with which you make machines learn, Deep Learning is a subset of some Machine Learning algorithms which are neuro network based.
  • Traditionally, neuro network means that there is only one hidden layer that is there between input and output. That is the standard neuro network. A specific set of algorithms which have more than one hidden layer, making the neural network deep, are called neuro network architectures. The entire field which deals with these neuro network architectures is called Deep Learning.
  • Deep Learning algorithms are extremely good at learning tasks and humans do not have to do a lot of work. For example, in feature engineering, trying to find out what are the right feature such that the machine can learn well. This usually works when you can provide it raw data sets. The algorithm will then itself work out how to perform a certain task.
• Machine Learning Vs AI
  Artificial Intelligence or AI is a super set of Machine Learning. Essentially, AI is any technique which makes machines intelligent. Any system which can ably perform tasks like a human is Artificial Intelligence while Machine Learning is a technique to make machines learn.

Typical Day of a Machine Learning Engineer

• Building models to be of useful assistance to the users or people. Understanding user preferences and needs so as to respond to them in a very personalized way.
• Writing code which takes the model to production. The coding has to be done keeping various parameters and the end user in mind. Engineering related nuances like faster response to user queries, scalability to a wider user base, etc. Specifically, it involves some level of modelling work and also a certain level of engineering work.

Corporate Culture Of Facebook

• Culture of open communication. Every level of employee is encouraged to talk to people, colleagues or peers. They can ask questions without being made to feel uncomfortable or guilty about it. One is free to communicate, voice their opinions and engage in a rational dialogue even with the seniors and the management cadre. This imparts a sense of belonging to every employee there.
• Culture of giving ownership to every single employee of the company. Facebook treats every employee as the owner of the company and instils in them the pride and joy on company achievements and a feeling of responsibility when the company messes up. Every employee there is a team player and feels a sense of ownership in the company.