Machine Learning

MACHINE LEARNING

Machine learning is a division of artificial intelligence(AI) that concentrate on the development of computer programs. It based on the idea that systems can learn and develop from access data, observations or experience without being expressly programmed.

The primary aim for machine learning is tooffer systems the ability to automate data-driven model building and make decisions with minimal human intervention.

Machine Learning – Algorithms That Generate Algorithms

Algorithms are a series of command used to instruct computers in new tasks to solve problems. Computer algorithms based on particular instructions and rules to organize tremendous amounts of data into intelligence and services.

Provide training data to a learning algorithm is the basic practice of machine learning. Based on the implication from the data, the learning algorithm can generate a rule set.It is fundamentally generating a new algorithm as a machine learning model.

By using different training material (data or experience) with same learning algorithm, it could be used to generate different data models.

• Supervised Machine Learning Algorithms:
  – The training material is about the analysis of a known training dataset
  – Produces an advanced function to predict the output values
  – Use labelled examples to apply learning outcomefor predicting future events
  – Able to provide targets for any new input after adequate training
  – Compare its output with the intended output
  – Find errors to modify the model accordingly
  
  
• Unsupervised Machine Learning Algorithms:
  – The training material is unclassified or unlabelled data
  – Infer a function to describe a hidden structure from unlabelled data by unsupervised learning studies
  – Cannot figure out the right output
  – Analyse the data and make conclusion from datasets to define hidden structures from unlabelled data
  
  
• Semi-Supervised Machine Learning Algorithms:
  – Between supervised and unsupervised learning
  – The training material is labelled and unlabelled data
  – Provide relevant training and resources for acquired labelled data
  
  
• Reinforcement machine learning algorithms:
  – Produce actions and discover error / reward to interact with its environment
  – The reinforcement learning is to have trial and error search and delayed reward
  – To maximize the performance, determining the best behaviour within unidentified context automatically is necessary
  – The reinforcement signal is to learn which action is the best by provided simple reward feedback or punishments as signals to agent
  
  

Machine learning requires extra time and resources to train it properly to enable analysis of massive quantities of data. After the learning processes, it can identify profitable opportunities or dangerous risks from more accurate results rapidly.To be more effective on processing enormous volumes of information, machine learning with AI and cognitive computing should combine together.

How Machines Learn

Machine learning methods can be categorized as three general types:

• Supervised Learning:
  – The learning algorithm is given labelled data and the desired output.
  
  
• Unsupervised Learning:
  – The unlabelled data given to the learning algorithm is to identify data patterns.
  
  
• Reinforcement Learning:
  – Dynamic environment provides rewards and punishments to interact with algorithm.
  
  

Key Considerations

To address AI to the context of people’s trust in the internet, there are several specific factors that must be considered:

• Socio-economic impacts
• Transparency, bias and accountability
• New uses for data
• Security and safety
• Ethics
• New ecosystems