Learn Before
Two Design Choices in a Modular Machine Learning Pipeline
Question: When building a machine learning system from separate stages rather than one fully integrated model, what two design choices must the practitioner make?
Sample answer: The practitioner must decide which stages belong in the system, and then decide the interface between those stages. In other words, they choose the pipeline components and how data or predictions move from one component to the next.
Key points:
- Choose the stages or modules in the pipeline.
- Define how one stage passes output to the next.
- Describe a modular, non-end-to-end system.
Rubric: The essay should clearly identify both the choice of pipeline stages and the choice of how those stages connect or hand off information.
0
1
Tags
Machine Learning
Deep Learning
Supervised Learning
Dive into Deep Learning @ D2L
Data Science
Machine Learning Strategy
Machine Learning Yearning @ DeepLearning.AI
Related
Use an Intermediate Detector When It Simplifies the Next Step
In a modular machine learning pipeline, what two design choices must be made first?
If a team does not build an end-to-end model, it no longer needs to decide how the system is divided into stages.
When a machine-learning task is split into several stages, you must design the _____ and decide how each stage passes information to the next.
Match each system-design term to its description.
Order the steps for designing a machine learning system that is not end-to-end.
In a multi-stage machine learning system, what does it mean for one stage to 'feed into' the next stage?
True or False: When you design a pipeline that is not end-to-end, you must choose both the individual stages and the way information moves from one stage to the next.
When assembling a multi-step workflow, you still need to decide how the pieces should _____ together.
Match each pipeline design scenario to the decision it illustrates.
Order the steps for deciding whether to build a modular machine learning pipeline.
Two Design Choices in a Modular Machine Learning Pipeline
Designing a Modular ML Workflow
Key Design Choices in a Modular ML System