1. Knowledge Base
  2. Getting Started with Levity

The Levity glossary

AI, ML, and Levity app terminology explained

What is a/an...

  • AI Block: Your custom machine learning model based on the training data you have trained it with.

  • API: Application Programming Interface - a method for software programs to communicate with each other using code. Learn more about APIs.
  • Classification: Sorting data points into different buckets. Read more about the subject here.
  • Confidence: Much of machine learning involves estimating the performance of a machine learning algorithm on unseen data. Calculating the confidence is a way of quantifying the uncertainty of such a prediction.
  • Error rate: We keep some amount of training data for testing and at the end of the training process, we automatically measure how many errors the model makes per 100 predictions. Read more about error handling here.
  • Flow: We have a visual workflow builder that lets you connect different applications, AI Blocks, and determine the right actions.
  • HITL: Human in the Loop - adding a human helping hand into the training and feedback of the model, to increase accuracy over time. Read more on this topic here.
  • Human review: Our platform comes with the ability to give feedback on predictions that the model is unsure about.
  • Integrations: We have developed direct connections with various 3rd-party applications that allow you to move data and actions between them in a seamless fashion. We are adding new ones on a continuous basis and you can find the complete overview here.
  • Label: Also known as a tag, category, target, or class. It is what the model will provide to you for unseen data.

  • Performance: Simply put, how well the AI Block does based on the information it has. Learn more about how we measure performance.
  • Prediction: What the AI Block believes to be the right label for an unseen data point.
  • Trigger: Every Flow requires a start signal. This can be a new file, an email attachment, or any other event that you determine to be the starting point of the process.
  • Unlabeled data: Any data that hasn't been assigned a category