Enterprise Technology Terminology: M

320 results · page 12 of 16

• Model Lifecycle Governance

  Model lifecycle governance is a formal framework of policies, processes, and controls that directs how organizations develop, validate, deploy, monitor, and retire machine learning and AI models, ensuring traceability, accountability, and compliance across model use in enterprise environments.

• Model Management System

  Model Management System is an enterprise software and process framework that centralizes how organizations register, version, deploy, monitor, and govern analytics, machine learning, and AI models across environments, supporting traceability, risk management, and controlled, repeatable operations for model lifecycle management.

• Model Monitoring

  Model monitoring is the continuous observation and measurement of machine learning models in production to track performance, data quality, and risk indicators against defined baselines, supporting MLOps, governance, compliance, and operational decisions across enterprise AI and analytics workloads.

• Model Optimization

  Model optimization is the process of refining trained machine learning or AI models and their execution environment to reduce compute, latency, and memory usage while preserving required accuracy, enabling reliable, cost-controlled deployment across data center, cloud, and edge environments in enterprise settings.

• Model Orchestration Engine

  Model orchestration engine is a software capability that schedules and coordinates the execution of machine learning and AI models across pipelines and infrastructure, enabling repeatable, governed, and auditable model operations within enterprise data, MLOps, and application architectures.

• Model Overfitting

  Model overfitting occurs when a machine learning model learns noise and artifacts from training data, achieving low training error but poor performance on new data. It matters in enterprises because it undermines prediction reliability, model risk management, and production deployment quality.

• Model Parallelism

  Model parallelism is a distributed training technique that partitions a single machine learning model across multiple devices so larger models can run within hardware limits, which matters for enterprises building and operating large-scale AI workloads on shared compute infrastructure.

• Model Parallelism Engine

  Model parallelism engine is a software layer that partitions a neural network across multiple devices and coordinates distributed execution, allowing enterprises to train and run very large AI models within hardware, memory capacity, and operational constraints of their infrastructure.

• Model Partitioning Strategy

  Model partitioning strategy is a planned approach for dividing an AI or machine learning model across multiple hardware or processes to support scalability, performance, security, and governance requirements in enterprise environments and distributed computing architectures.

• Model Poisoning

  Model poisoning is an adversarial machine learning attack where an attacker corrupts training data, model updates, or training workflows so a deployed model embeds attacker-chosen behavior while still appearing to perform acceptably in enterprise applications and risk-managed AI architectures.

• Model Provenance

  Model provenance is the documented record of an AI or machine learning model’s origin, data lineage, training process, and version history, which enterprises use to support governance, auditability, regulatory compliance, and lifecycle management of models deployed in production systems.

• Model Pruning

  Model pruning is a model compression technique that removes parameters or structures from trained neural networks to reduce size and computation, enabling deployment on constrained or cost-sensitive infrastructure while maintaining accuracy levels that meet enterprise performance and service objectives.

• Model Quantization

  Model quantization is a neural network compression technique that encodes weights and sometimes activations in lower-precision numeric formats to reduce memory, compute, and energy usage, which supports cost-efficient and latency-aware deployment of machine learning models in enterprise environments.

• Model Registry

  Model registry is a centralized system of record for storing, versioning, and governing machine learning and AI models and their metadata, enabling enterprises to control model lifecycle, support reproducibility, and align model deployment with operational, risk, and compliance requirements.

• Model Retraining

  Model retraining is the process of updating an already deployed or validated machine learning model with new or revised data so its performance remains aligned with current conditions, supporting reliability, compliance, and risk control in enterprise AI systems.

• Model Retraining Schedule

  Model retraining schedule is a documented plan that defines when and under which monitored conditions an enterprise updates a machine learning model with new data, ensuring controlled lifecycle management, traceable changes, and alignment with governance, risk, and compliance requirements.

• Model Risk Management

  Model risk management is the governance, processes, and controls organizations use to identify, assess, and mitigate risks from models used in decision-making, helping enterprises manage financial, compliance, and operational exposure associated with statistical, machine learning, and other analytical models.

• Model Rollback Policy

  Model rollback policy is a documented set of rules and procedures that governs when and how an enterprise reverts a deployed AI or machine learning model to a previous or fallback version, supporting operational continuity, governance, and auditability.

• Model Safety Envelope

  Model safety envelope is a defined set of technical, operational, and policy boundaries within which an AI or machine learning model may operate, used by enterprises to align model behavior with documented safety, reliability, and risk management requirements.

• Model Serving

  Model serving is the process and infrastructure that deploy trained machine learning models into production and expose them via stable interfaces for inference, enabling enterprises to integrate AI predictions into applications under managed performance, reliability, security, and governance controls.