Retraining
Model retraining is the process of training an existing machine-learning model again, typically with new or updated data or changed training conditions.
Machine Learning System Design
Retraining is the process of training a machine learning model again, typically with changed data, initialization, or training conditions.
- How it is used
- Guides decisions about updating a deployed model when data changes, while keeping its behavior reliable across releases.
- How deep it goes
- Provides monitoring and investigation steps before retraining, but not a retraining algorithm or implementation procedure.
- Examples
- Uses weekly batch updates and an antifraud model whose classifications fluctuate across releases.
- What you can do after
- Evaluate/understand whether drift warrants retraining and whether a new release changes predictions acceptably.
Read how Machine Learning System Design explains retraining →
Machine Learning Platform Engineering
Model retraining is the process of training a machine-learning model again using newly available or updated data.
- How it is used
- Keeps production models effective by responding to drift, performance degradation, new data, and changing business conditions.
- How deep it goes
- Explains triggers, automation, validation, monitoring, and deployment integration, but does not provide implementation code or detailed algorithms.
- Examples
- Uses monthly schedules, loan-approval thresholds, identity-card detection, recommender updates, and candidate architecture evaluation.
- What you can do after
- Operate/configure automated retraining pipelines that respond to monitored conditions and deploy validated model replacements.
Read how Machine Learning Platform Engineering explains retraining →
Outlier Detection in Python
Retraining is the process of fitting a model again using new, updated, or expanded training data.
- How it is used
- Maintains outlier-detection performance as data, goals, and labeled evidence change over time.
- How deep it goes
- Explains retraining choices and regression checks, but does not detail model-fitting algorithms or optimization internals.
- Examples
- Uses recent data, labeled records, synthetic data, and alternative detector configurations as retraining and testing artifacts.
- What you can do after
- Diagnose retraining regressions by checking known outliers and non-outliers against stored labeled records.
Financial AI in Practice
Model retraining is the process of training an existing model again, typically with updated data, to refresh its learned parameters.
- How it is used
- Keeps financial models responsive to changing conditions and incorporates new training data or investigation feedback.
- How deep it goes
- Details a full-data credit-model fit and fraud retraining controls, while reserving final regulatory validation for an isolated hold-out set.
- Examples
- Uses a WoE-coded credit dataset and fraud-investigation outcomes as retraining examples.
- What you can do after
- Apply retraining by fitting an XGBoost classifier on the full WoE-coded training dataset after validating hyperparameters.
Related concepts
- driftaddresses · Drift is a change over time in a system’s behavior, state, data, or configuration relative to a reference or expected baseline.
- monitoringused by · Monitoring is the ongoing observation and measurement of a system’s behavior, state, or operational signals to assess its condition and detect changes or problems.
- fine-tuningalternative to · Fine-tuning is adapting a pretrained machine-learning model by training it further on targeted data or objectives for a task, domain, or behavior.
- checkpointingused by · A checkpoint is a designated point in a process at which its state or progress is recorded or evaluated, often to support recovery, continuation, or control.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.