Difference between supervised learning and unsupervised learning
Supervised Learning: When there is a person who tests and decides whether you have gotten the answer correctly as the student is learning a specific task is seen as supervision. Similarly, when you train an algorithm, the idea of supervised learning deals with providing a full collection of labeled data.
Completely labeled ensures that the response the algorithm can come up with on its own is labelled with each example in the training dataset. So, a classified flower image dataset will tell the model which pictures were of roses, daisies and daffodils. When shown a new image, the model compares it to the training examples to predict the correct label.
You want to train a computer, for example , to help you predict how long it will take you from your office to drive home. You begin by creating a collection of labeled data here. This data contains
- Climate condition
- Time of the day
- Holiday
Unsupervised Learning: Unsupervised learning is a methodology for machine learning, where the model does not need to be supervised. Instead, to discover data, you need to activate the model to operate on its own. It deals primarily with unlabelled information.
Take the example of a girl and her family dog.
She recognizes this dog and names it. A few weeks later, a friend of the family brought a dog along and tried to play with the boy.
The baby hadn’t seen this dog before. But certain traits (2 ears, eyes, walking on 4 legs) are known as being like her pet dog. Like a puppy, she will recognise a new animal. This is unsupervised learning, where you are not trained, but you learn from the data (a dog’s knowledge in this case). If this learning had been supervised, the family friend would have told the baby that it was a dog.
Deep Learning
Deep learning is a technique of machine learning that teaches computers to do what comes to humans naturally: to learn by example. A main technology behind driverless cars is deep learning, which enables them to identify a stop sign or to differentiate a pedestrian from a lamp post. It is the key to voice control in consumer devices like phones, tablets, TVs, and hands-free speakers. Deep learning is getting lots of attention lately and for good reason. It’s achieving results that were not possible before.
In deep learning, a computer model learns to perform classification tasks directly from images, text, or sound. Deep learning models can achieve state-of-the-art accuracy, sometimes exceeding human-level performance. Models are trained by using a large set of labeled data and neural network architectures that contain many layers.
Reinforcement learning
The training of machine learning models to make a series of choices is reinforcement learning. In an unpredictable, potentially complex environment, the agent learns to attain a target. Artificial intelligence faces a game-like scenario in reinforcement learning. To come up with a solution to the problem, the machine uses trial and error. For the actions it does, the artificial intelligence earns either incentives or punishments to get the system to do what the programmer wants. Its objective is to maximize the overall reward.
In normal circumstances, for example, we will need an autonomous vehicle to put safety first, decrease travel time , reduce emissions, provide comfort to passengers and comply with the law. On the other hand, we will emphasize speed in an autonomous racing car even more than the comfort of the driver. The programmer is unable to predict anything that could happen on the path. The programmer trains the reinforcement learning agent to be able to learn from the structure of rewards and penalties instead of constructing long “if-then” instructions. The agent (another name for the task-performing reinforcement learning algorithms) gets rewards for achieving particular goals.
Ontology
An ontology describes a set of representational primitives from which to model a domain of knowledge or discourse in the sense of computer sciences and information sciences. Typically, the representational primitives are classes (or sets), attributes (or properties), and relations (or relationships between members of the class). Details on their significance and restrictions on their logically consistent implementation are included in the descriptions of representational primitives. Ontology can be seen as a degree of abstraction of data structures in the sense of database systems, similar to hierarchical and relational models, but intended to model information about persons, their characteristics, and their relationships with other individuals. Usually, ontologies are defined in languages that allow abstraction away from data structures and execution strategies; in practice, ontology languages are closer to first-order logic in expressive power than languages used to model databases.
Decision Tree
Decision Trees are great tools to help you select from various courses of action.
They have a highly efficient framework within which options can be set out and the potential consequences of selecting such options can be examined. They also allow you to build a realistic image of each potential course of action’s risks and rewards.
2020-11-4-1604483298