What is Machine Learning?(part 1)

Photo by Owen Beard on Unsplash

• Definition
• Types of Machine Learning
• Difference between Artificial Intelligence, Machine Learning and Deep Learning
• A list of top 10 Machine Learning Algorithms
• Deep Learning and Deep Reinforcement Learning
• Drawbacks in Machine Learning
• Conclusion
• References

Definition

Machine learning is a technique of training a machine to perform activities of the human brain in a faster and more accurate manner than an average human being will. It uses statistical methods to enable machines learn and improve with experience the same way humans learn and improve over time.

Types of Machine Learning Algorithms

1. Supervised Leaning
2. Unsupervised Learning
3. Reinforcement Learning

Supervised Learning

In Supervised Learning, a function is generated to map inputs to desired outputs using set of variables. The inputs and outputs are supplied thus allowing the algorithm to study the relationship between the two(2) patterns so that similar patterns that count in future will be able to be predicted.

The machine learns under supervision like a student in creche with a kind of labelled data. It is this concept that google uses to classify emails intelligently into spam and non spam. If you have a clearer and more labelled data, then opt for Supervised Learning.

Unsupervised Learning

There is no target or outcome variable to predict in Unsupervised Learning. The algorithm studies the input supplied and classifies them into clusters. No output is given, thus the machine does not learn under supervision just like a student in the University. You can opt for Unsupervised Machine Learning with larger and unlabelled datasets.

Unsupervised Learning has shown a great success in many modern Artificial Intelligence Applications like face detection and object detection.

Reinforcement Learning

In Reinforcement Learning, the machine is trained to make specific decisions(to establish or encourage a certain behavior) on its own. This technique was initially developed for machines to learn how to play games. A machine is exposed to an environment where it trains itself continuously using trial and error. For each right move, the machine is rewarded but penalized for a wrong move. It is with this concept that the game of chess and several other games were programmed. These games were made to play against themselves a million times by rewarding the right moves and allowing the machines get accustomed to them.

Reinforcement Learning is more concerned with exploring better solutions versus exploiting the learned solutions. It learns from past experiences and tries to capture the best possible knowledge to make accurate decisions.

Difference between Artificial Intelligence, Machine Learning and Deep Learning

Artificial Intelligence is a technique that enables the machine to imitate human behavior.

Machine Learning is a subset of Artificial Intelligence. It is supplying an input to an algorithm and training that algorithm. The more accurate the input, the more accurate the algorithm.

Deep Learning is a subset of Machine Learning inspired based on the concept of the brain cells called neurons which led to the concept of Artificial Neural Networks(ANN). It does not perform well with little amount of data like Machine Learning.

Artificial Intelligence >> Machine Learning >> Deep Learning

A list of top 10 Machine Learning Algorithms:

1. Linear Regression
2. Logistic Regression
3. CART(Classification and Regression Trees)
4. Naive Bayes
5. KNN(K-Nearest Neighbors)
6. SVM(Support Vector Machine)
7. K-means Algorithm
8. PCA (Principal Component Analysis)
9. Random Forest
10. Gradient Boost and AdaBoost

The algorithms above fall under the three(3) broad categories of Machine Learning algorithms. They are subsets of Supervised, Unsupervised and Reinforcement Learning.

Deep Learning and Deep Reinforcement Learning

Deep Learning

Deep Learning consists of Deep Neural Networks, Deep Belief Networks, Recurrent Neural Networks and Convolutional Neural Networks which have been successfully applied in solving problems of Computer Vision, Speech Recognition, Natural Language Processing, games, Drug Design, Medical Image Analysis, e. t. c

There is a demand for huge amounts of data and huge processing data which is usually available these days. Deep Learning performs better with large data.

Deep Reinforcement Learning

Deep Reinforcement Learning combines the technique of both Deep Learning and Reinforcement Learning. Here, Reinforcement Learning algorithms like Q-Learning are combined with Deep Learning to create powerful Deep Reinforcement Learning algorithms. This technique has demonstrated a great success in Robotics, video games, finance, health care, e. t. c.

Some Applications of Machine Learning

• Mobile Apps(For organizing photos, face detection, face ID, face tagging, image identification)
• Spam Filtering
• Credit Card Fraud Detection
• Digit recognition on checks, zip codes
• MRI Image Analysis
• Recommendation System
• Search Engines
• Handwriting Recognition
• Scene Classification
• Self Driving Cars
• Speech Recognition(Apple’s Siri, Amazon’s Alexa, Microsoft’s Cortena, Google’s Assistant.

Drawbacks of Machine Learning

Some of the drawbacks or disadvantages of Machine Learning include:

• Machine Learning is Susceptible to errors. There is no perfect level of accuracy attained yet, so errors are bound to occur during predictions.
• It is time and resources consuming. Algorithms need so much time to learn and develop on their own. Moreover, successful training of Deep Neural Networks may require several weeks of training time.
• Scarcity of data. Since Deep Learning requires humongous data to be able to function well, the shortage of data can disrupt the function of models. The more the data, the more accurate the models and vise versa.

Conclusion

Photo by Devin Avery on Unsplash

Machine Learning is a very important aspect to dive into as an individual who embraces technology or is curious about the future. It requires skills like Statistics, Probability theories, Calculus, Optimization techniques, Linear Algebra, Visualization and so on.
Because it is a broad, it should be studied, one topic at a time as it is one of the fastest growing technologies of our time. Businesses, Jobs and Companies are being automated, thus the need for Machine Learning is arising.

References

https://www.youtube.com/watch?v=WSbgixdC9g8

www.Udacity.com(Introduction to Machine Learning)

www.researchgate.net(Broadly Machine Learning)

www.Edureka.com(Supervised and Unsupervised Machine Learning)

Tutorials Point(Machine Learning Tutorial)