Machine Learning -Basic Terminologies in Context

Machine Learning – Basic Terminology in Context – Machine learning should be treated as a culture in an organisation where business teams, managers, and executives should have some basic knowledge of ML and its terminology. There are many online courses available that are designed for students, employees with little or no experience, managers, professionals, and executives to give them a better understanding. This post is part 2 of Machine Learning (ML) – Basics you need to Know.

• Machine Learning – Introduction to Its Algorithms
• How Machine Learning Algorithms Works
• Machine Learning as a Service

Machine Learning Outlook

Today’s machines are learning and performing tasks; that was only done by humans in the past like making better judgment, decisions, playing games, etc. This is possible because machines can now analyse and read through patterns and remember learnings for future use. How to harness this magnificent machine learning and its bundle pack in real-life business is still a challenge for many though. Why ML is so good today; for this, there are a couple of reasons below but not limited to though.

• The explosion of big data
• Hunger for new business and revenue streams in this business shrinking times
• Advancements in machine learning algorithms
• Development of extremely powerful machine with high capacity & faster computing ability
• Storage capacity

The primary goal of machine learning (ml) is to build an automated data model for analytical reasons. The objective behind the goal is to build a system that learns from the data based on the applied algorithm. The output can be obtained by mapping output to input or detecting patterns/structure or learning by reward/punishment method.

Not to be confused AI is not Machine learning though Machine Learning is a technique out of the AI bundle to achieve artificial intelligence. Let’s just say AI generally defines & demonstrate creativity in the following traits:

• Planning & Predicting
• Learning & Adopting
• Reasoning & Logic Building
• Problem Solving & Avoiding
• Knowledge Representation
• Perception & Reasoning
• Motion Detection and Manipulation

Machine Learning helps to implement all the above by using correct methods and algorithms with correct data sets. In earlier times it was believed that human intelligence can be precisely described, and machines can simulate it with AI. Before the machine starts attempting simulation, it needs to do learning with lots of data.

Machine Learning – Integral Issues

Machine learning techniques are accelerating almost daily to bring good values to the businesses of today. It is a distinct game. It covers a huge and extensive territory at the same time it has impacted many almost every business domain/vertical.

Machine learning today can tackle tasks in language, video, image processing, anomaly detection, pattern recognition, credit scoring, sentiment analysis, etc. This whole game and environment are entirely centered around artificial neural networks working like the human brain and machine learning.

The Machine Learning hype, too much information on the internet, and using ML terms by almost every tech show have actually created a “misinformation epidemic” of ML.  It is revolutionizing the way we do our business and what should be done to improve upon it. ML develops its own encompassing strategy from the experience it comes across over the period. Mathematics, statistics, programming, and the common sense of human beings are now part of integral components of machine learning.

The common difference between machine learning techniques has both sides of the coin the disadvantage of supervised machines is their need for large sums of labeled data. Labeling a large number of samples is costly and time-consuming. Unsupervised learners don’t have this disadvantage, but they tend to be less accurate. Naturally, there is a strong motivation to improve unsupervised machines and to lessen the reliance on supervised ones.

You can view GANs and RLs as means of improving unsupervised machines (neural networks). The strategy is learning from the experience it comes across over a period of time.

The correct information and little basic information about machine learning for non-tech and tech business professionals is the need of today. For business people, knowledge in the area of machine learning is less important than knowledge of their own business to create required and correct data models.

Machine Learning gives exposure to businesses to make data-driven, more informed, and intelligent decisions after inputting correct data from the business and not by creating complex and unnecessary use of relevant algorithms. Such decisions help to make faster and better results compared to traditional approaches. There are a lot of common mistakes that are made and should be avoided to successfully submerge machine learning in an analytics strategy for better business transformation. Some classic issues are 

Inadequate Infrastructure – As mentioned above the requirement of a powerful machine and accelerated hardware with high storage capacity and faster computing ability is a basic need in machine learning which is normally ignored. This is a fatal problem and issue.

Data Quality Problems – The explosion of big data has created great hunger for new business and revenue streams in these shrinking times. What data is correct and what is not is the primary issue to resolve with a business goal. For example, an insurance company selling tooth insurance needs data on how teeth hygiene are maintained by customers along with brushing habit and brand of toothpaste used, time spent on brushing teeth, and frequency rather than data on blood report and blood pressure.

Correct skilled resources – Implementing machine learning without qualified data scientists is the biggest issue of any business today. The cost of the skilled resource is pretty high, supply is very low and demand is big.

Implementation without Strategy – Advancements in machine learning algorithms have created a tsunami-like environment where businesses are running behind this buzzword.  Unfortunately, the challenge is big as not all corporates can make a decision about which strategy to be picked up at what time at this time of digital transformation. Not all business models need complex machine learning without proper analysis. Should machine learning be brought in the form of advancement to analytics or simple regression models are enough.

• Top Machine Learning Algorithms – Data Scientist Basic Tool Kit
• The Exciting Evolution of Machine Learning
• Astonishing Hierarchy of Machine Learning Needs

AA

Machine Learning – Common Terms

Learning in simple words can be defined as “Learning, validating, testing and repeating” for every new subject. In Machines, it’s more about the accuracy of the improvement of the algorithms over time. Machine Learning accuracy improves over time the more data it eats better it gets better. We have machine learning all around us like in our email box, eCommerce site, and even at banks. The fundamental terminologies in ML often overlap with statistics but that is just limited to the basic and initial part only. How learning evolves and at what rate and how it over right its previously known learnings are interesting items to look at.

• Learning  – As mentioned above learning in simple words can be defined as “Learn, validate, test, and repeat” for every new subject. The 3 key factors in machine learning are
  • Learning Rate –  Also called step size, LR is a kind of fine-tuning parameter used for moving towards a minimum of a loss function in the optimization algorithm. In Neural networks, it’s called a value factor to determine the level of updates for the value of weights of the network.
  • Learning Rate Schedule – Usually a constant learning rate. In neural networks, it is used for training to achieve optima of neural network performance. LRS helps to reduce the learning rate during training by adjusting to a pre-defined schedule. The best approach to learning rate schedules should be to decrease the learning rate linearly from a bigger value to a small value.
  • Learning Rate Decay – Movement during the process of achieving a local minimum in gradient descent that helps to speed up the learning algorithm. By slowly reducing the learning rate over time we achieve learning rate decay. Initially, the learning rate is bigger which reduces to a smaller value this process comes under learning rate decay.

• Gradient –  A simple vector that provides direction for the maximum rate of change to reach the optimal solution.  If walking from the top of the hill to the bottom then the steepest direction would be the fastest, optimal, and logical. Gradient provides the steepest direction. In the machine, the learning goal is always to achieve optimal solutions. The gradient is a vector that provides directions for the maximum rate of change to reach the optimal solution.
  • Vanishing Gradient – Vanishing gradient is a problem during the training phase of neural networks. During the backpropagation process instability of gradient values, it causes instability that affects the earlier layers within a neural network. A big limiting factor for the performance and accuracy of a neural network.
  • Exploding Gradient – Pretty much similar to the vanishing gradient problem, the exploding gradient problem is a result of the instability of the gradients within a neural network during backpropagation.
  • Gradient Clipping –A technique that is used to bring regularity in the instability of gradient values within a neural network. This is achieved by enforcing a threshold on the values that the gradients can take.

The fundamental terminologies in ML algorithms are used in the majority with teaching intentions while deep-learning procedures are tweaked to learn themselves. ML often overlaps with statistics but that is just limited to the basic and initial part only. The learning algorithm normally has three stages i.e. representation, evaluation, and optimisation. Reaching optimal accuracy and precision is always the dream.

Points to Note:

All credits if any remains on the original contributor only. We have covered all basics around Machine Learning. Machine Learning is all about data, computing power and algorithms to look for information. In the previous post, we covered Generative Adversarial Networks. A family of artificial neural networks.

Books + Other readings Referred

• Research through open internet, news portals, white papers and imparted knowledge via live conferences & lectures.
• Lab and hands-on experience of  @AILabPage (Self-taught learners group) members.

Feedback & Further Question

Do you have any questions about Supervised Learning or Machine Learning? Leave a comment or ask your question via email. Will try my best to answer it.

Conclusion –  I particularly think that getting to know the types of machine learning algorithms actually helps to see a somewhat clear picture. The answer to the question “What machine learning algorithm should I use?” is always “It depends.” It depends on the size, quality, and nature of the data. Also, what is the objective/motive data torturing?

As more we torture data more useful information comes out. It depends on how the math of the algorithm was translated into instructions for the computer you are using. And it depends on how much time you have. To us, at AILabPage we say machine learning is crystal clear and ice cream eating task. It is not only for PhDs aspirants but it’s for you, us and everyone.

============================ About the Author =======================

Read about Author at : About Me

Thank you all, for spending your time reading this post. Please share your opinion / comments / critics / agreements or disagreement. Remark for more details about posts, subjects and relevance please read the disclaimer.

FacebookPage    ContactMe      Twitter

====================================================================