Machine Learning
Machine Learning is a subset of Artificial Intelligence that inherits learning capability without being programmed to do so. ML has the ability to enhance simple automation processes, with the power to predict the outcome of an event quite accurately. ML has been here for several decades now. It is a data-hungry application, that requires high computational power. Due to the explosion of data in recent years and the promise of exponential data growth, ML is now more popular than ever!
While the world today is saturated by AI, ML applications, it has become overtly important to understand and identify the types of machine learning that we may encounter or just to get the pulse of the world we live in!
Types of ML are as follows:
- Supervised Learning
- Unsupervised Learning
- Reinforcement Learning
Supervised and unsupervised learning are two essential methods through which machines can automatically learn and improve from experience. The learning process begins with lots of observational data (such as examples or instructions if you will) with the end target to seek patterns. Those patterns are then used to forecast decisions using the input examples/ instruction that we feed into the machine.
- Supervised Learning
In supervised learning, people train the machine using labeled data. Here, labeled data means it is already tagged with the right answer. It is a task-oriented model that requires more and more input examples, till it learns to predict the output accurately.
In mathematical terms, when given a set of data points {x^(1), ..., x^(m)} is associated with a set of outcomes {y^(1), ..., y^(m)}, we want to build a classifier that learns how to predict y from x.
Supervised learning models are designed to work quickly, with the power of nearly limitless resources. The algorithms are so minimal and clean, the susceptibility to errors is very low. It provides the greatest anomaly detection algorithms.
Some examples of supervised learning applications include:
- Credit card fraud detection (fraud, not fraud)
- Email spam detection (spam, not spam)
- Text sentiment analysis (happy, not happy).
- Patient risk (such as high-risk patient, low-risk patient) or for predicting the probability of congestive heart failure.
2. Unsupervised Learning
Unsupervised learning is a type of algorithm that does not require labels to be given to the algorithm. It learns patterns from untagged data.
The goal of unsupervised learning is to find hidden patterns in unlabeled data :{x^(1),...,x^(m)}
The algorithm is fed a lot of data and given the tools to understand the properties of the data. It then learns to group, cluster, and organize the data. In other words, unsupervised learning is an intelligent algorithm that can take terabytes of unlabeled data and make sense of it.
Some examples of unsupervised learning applications include:
- Social network analysis
- Image Segmentation
- Handwriting recognition
- Anomaly detection and etc.
3. Reinforcement Learning
Reinforcement learning is a closed-loop model that enables an agent to learn in an interactive environment by trial and error using a feedback loop. RL requires a lot of data, therefore it is most applicable in domains where simulated data is readily available like gameplay, robotics.
Both supervised and reinforcement learning use the mapping between input and output. Unlike supervised learning where feedback provided to the agent is the correct set of actions for performing a task, reinforcement learning uses rewards and punishment as signals for positive and negative behavior.
As compared to unsupervised learning, reinforcement learning is different in terms of goals. While the goal in unsupervised learning is to find similarities and differences between data points, in reinforcement learning the goal is to find a suitable action model that would maximize the total cumulative reward of the agent.
Some examples of reinforcement learning are as follows:
- AI based computer games eg. AlphaGo Zero, ATARI games, Backgammon, etc
- Robotics and industrial automation eg. DeepMind’s work on Deep Reinforcement Learning for Robotic Manipulation with Asynchronous Policy updates is a good example of the same.
- Text summarization engines, dialog agents (text, speech)
While the field of Machine Learning can provide a lot of advantages and it has had such a profound impact on the world, common folks have a hard time grasping its full potential and capabilities, and perhaps more importantly, its limitations. As we know now, that ML is a data-driven application, it can and is facing some ethical boundaries while being trained with huge amounts of data. For instance, for a self-driving car to run every day, it will generate 4000Gb of data in a day alone. You can certainly imagine the amount of computational resources required to be fed into the intelligent machines.
Machine learning and artificial intelligence will continue to revolutionize the industry and will only become more prevalent in the coming years. Whilst we will mostly be the ones utilizing ML and AI to their fullest extent, we also need to remember the limitations of the tools in our power use.
