General Terminology

Term description
Training data Training data is the data that represents the correct answer to the model’s objective. It is used as the input data when learning under supervised learning.
Model A product generalized through training by applying algorithms formulated for the training data. ABEJA Platform includes the entire program needed to make inferences. In this document, the term”model” means a model on the ABEJA Platform.
Learning Learning is the process of generating a model by using algorithms formulated for the training data.
Re-learning The process of relearning new training data to generate a model.
Inference The process of applying a model to input data to predict the desired continuous/discrete values.
GPU A processor used to rapidly calculate three-dimensional graphics. GPUs are often used in deep learning to process calculations.
Validation Validation is the ability to validate the correctness of formats, values etc. based on rules set for the input data.
Blue-Green Deployment Blue-Green Deployment is a technique that enables switching between idle and live versions

ABEJA Platform Terminology

Term Description
Users Refers to users of ABEJA Platform. Users must register and belong to an organization in order to use ABEJA Platform.
Organization AA unit that manages users and various resources on the ABEJA Platform. As a general rule, resource access is restricted to each organization.
Role Refers to the role assigned to a user on the ABEJA Platform. Executable permissions are determined based on this role.
Data source Data source refers to the source of data input to the ABEJA Platform. Examples include IoT devices or other systems that contain data.
Data lake AA feature for centrally managing a diverse range of data in various formats on the ABEJA Platform. All data input to ABEJA Platform from a data source is managed in a data lake.
Channel A function that manages input/output between a data source and data lake on ABEJA Platform. It is possible to limit, by channel, the data sources permitted to access the data lake.
Resource Objects identified by a unique ID within the ABEJA Platform, such as data sources, channels, models, etc.
API An HTTP Restful application interface for entering data into ABEJA Platform and retrieving executed results.
CLI A Command Line Interface for managing data and models executed on ABEJA Platform.
Management Console A graphical user interface that can be run on a web browser to access features available in ABEJA Platform.
Edge device management system A management feature mainly for deploying models learned on the ABEJA Platform to the edge of an IoT device etc.
Trigger A trigger is event-based execution function that applies a model as soon as data is saved to a data lake channel. A trigger can be used to specify an input channel for data and output channel for the executed result of a model.
Deploy To deploy a model trained in a local environment etc. so it can be run on ABEJA Platform.
Deployment Deployment refers to the individual deployment of a model that runs on ABEJA Platform and encompasses the broader concept that also includes the model version and service.
Version Version feature of a model deployed on ABEJA Platform. Multiple versions can be managed for each deployment.
Service The ability to, through HTTP, triggers, and batch files, run and manage the executable states of a specified version of a model deployed on ABEJA Platform.
Service endpoint For a registered service, the endpoint that runs a model that can be called externally using HTTP.
Edge A device etc. that make inferences closer to the target.
Daemon A process that resides in memory on Linux and UNIX to provide various services.