Organizing Model Experiments in Kubeflow Pipelines

PMLE By islever Editorial Last updated

Tracking and running ML experiments.

Short answer

D Create an experiment in Kubeflow Pipelines to organize multiple runs.

Question

You need to design a customized deep neural network in Keras that will predict customer purchases based on their purchase history. You want to explore model
performance using multiple model architectures, store training data, and be able to compare the evaluation metrics in the same dashboard. What should you do?
  1. Create multiple models using AutoML Tables.
  2. Automate multiple training runs using Cloud Composer.
  3. Run multiple training jobs on AI Platform with similar job names.
  4. Create an experiment in Kubeflow Pipelines to organize multiple runs. Correct Answer

Key Takeaways

The question tests the ability to identify the correct tool for experiment tracking and comparison, distinguishing between orchestration (Composer), managed ML (AutoML), and structured experiment management (Kubeflow/Vertex AI).

Learn how to track and compare multiple model architectures using Kubeflow Pipelines experiments for efficient ML workflow management.

Common mistake: Choosing Option C is common because AI Platform allows training, but it fails to address the need for centralized metric comparison and organization inherent in 'experiments'. Using 'similar job names' is also technically invalid as job names must be unique.

Expert Analysis

The question requires designing a workflow to explore multiple model architectures, store training data, and compare evaluation metrics in a single dashboard. The key constraint is the need for organized experimentation with custom Keras models.

Option A (AutoML Tables) is incorrect because it is a managed service for tabular data that automates model building. It does not allow for custom deep neural network architecture design or flexible hyperparameter tuning across different architectures.

Option B (Cloud Composer) is incorrect because it is an orchestration tool for managing workflows, not a dedicated experiment tracking or comparison platform. While it can trigger jobs, it lacks the native feature for comparing metrics and artifacts in a unified dashboard.

Option C (AI Platform with similar job names) is incorrect and misleading. First, you cannot use 'similar' job names; each job must have a unique name. Second, simply running jobs on AI Platform does not inherently provide a centralized dashboard for comparing experiments unless explicitly integrated with Vertex AI Experiments. The option describes a manual, unstructured approach that fails the 'compare in the same dashboard' requirement.

Option D (Kubeflow Pipelines experiment) is correct. Kubeflow Pipelines provides an 'Experiment' concept specifically designed to group related runs. This allows users to organize multiple training runs, log parameters and metrics, and visualize comparisons using tools like MLflow or Vertex AI's integration with Kubeflow. Although the modern Google Cloud answer would typically reference 'Vertex AI Experiments', among the provided options, Kubeflow Pipelines is the only framework that natively supports the concept of organizing and comparing experiments via its pipeline structure. The comment noting 'better in vertex AI during experiment phase only' highlights the evolution of the product, but D remains the best fit for the described functionality of organized experimentation.

Official Reference

https://www.kubeflow.org/docs/components/pipelines/concepts/experiment/

This link opens the official vendor documentation in a new tab.

Exam Strategy

When questions mention 'experimenting', 'comparing metrics', or 'tracking runs', look for options involving Experiment Tracking systems (like Vertex AI Experiments or Kubeflow Experiments). Avoid generic orchestration tools (Cloud Composer) or managed services (AutoML) that do not offer granular control over custom model structures and comparative analysis dashboards.

Frequently Asked Questions

Why is AutoML Tables not suitable for this scenario?

AutoML Tables automates model creation for tabular data but does not allow for custom deep neural network architecture design or flexible exploration of multiple custom architectures.

Can I use Cloud Composer for experiment tracking?

Cloud Composer orchestrates workflows but does not provide native experiment tracking or a unified dashboard for comparing model metrics. It triggers jobs but doesn't manage the experimental data itself.

More PMLE FAQ →

Community Votes

D
83%
C
17%

83% of anonymous learners picked answer D. Votes are pick records left by other test-takers — they are not the verified answer.

Discussion 21 contributions

Compiled from public community threads and reader submissions, lightly edited for clarity. Handles and vote tallies are shown as recorded, and comments reflect the views of individual contributors.

From public community threads 21

  1. ralf_cc 👍 14
    D - https://www.kubeflow.org/docs/about/use-cases/
  2. tavva_prudhvi 👍 8 Answered D
    The best approach is to create an experiment in Kubeflow Pipelines to organize multiple runs. Option A is incorrect because AutoML Tables is a managed machine learning service that automates the process of building machine learning models from tabular data. It does not provide the flexibility to customize the model architecture or explore multiple model architectures. Option B is incorrect because Cloud Composer is a managed workflow orchestration service that can be used to automate machine learning workflows. However, it does not provide the same level of flexibility or scalability as Kubeflow Pipelines. Option C is incorrect because running multiple training jobs on AI Platform with similar job names will not allow you to easily organize and compare the results.
  3. salsabilsf 👍 6
    Should be D
  4. Danny2021 👍 4
    D. In the new Vertex AI, it now supports experimentation with hyper parameter tuning.
  5. kfrd 👍 4
    C anyone? D seems to me like an overkill.
Show 16 more comments
  1. kaike_reis 👍 3
    (C) presents the most specific solution for what the question asks for: experimenting with models with their due comparisons. All of this is possible with the AI Platform. Furthermore, the question only speaks of experimentation. Kubeflow would be more powerfull if was a necessity for end-to-end pipeline.
  2. mmona19 👍 3 Answered D
    D- we need to use experiments feature to comapre models,having different jobnames is not going to help track experiments.
  3. tikka0804 👍 2
    I would vote for D but if C had said instead "different job names" .. would that have been a better option?
  4. sid515 👍 2
    C for me. It only talks about experimentation .. thats where AI platform fits better.
  5. mymy9418 👍 2 Answered C
    https://cloud.google.com/vertex-ai/docs/experiments/user-journey/uj-compare-models
  6. SamuelTsch 👍 2 Answered D
    D should be correct
  7. taylz876 👍 1 Answered D
    You want to try several architectures, keep the training data, and compare the evaluation metrics in one dashboard. That's what experiment tracking does: each training run is logged under one experiment, with its parameters and metrics, and you can compare the runs side by side. Kubeflow Pipelines has a built-in "experiment" feature for this. On the current exam, the matching tool is Agent Platform Experiments (formerly Vertex AI Experiments).
  8. NamitSehgal 👍 1 Answered C
    Similar job names is a bit of a confusion creator as we can not use same job names for sure. D sounds better but better in vertex AI during experiment phase only.
  9. Fatiy 👍 1 Answered D
    With Kubeflow Pipelines, you can create experiments that help you keep track of multiple training runs with different model architectures and hyperparameters.
  10. Mohamed_Mossad 👍 1 Answered D
    https://www.kubeflow.org/docs/components/pipelines/concepts/experiment/ https://www.kubeflow.org/docs/components/pipelines/concepts/run/
  11. tavva_prudhvi 👍 1
    How can we track the progress of each run and compare the results in the vertex AI dashboard?
  12. Liting 👍 1 Answered D
    C has similar job name, which make it wrong So correct answer should be D
  13. suresh_vn 👍 1
    D option C does not work since CAIP have updated to VertexAI
  14. Sum_Sum 👍 1 Answered D
    D - everything else is just nonsense
  15. PhilipKoku 👍 1 Answered D
    D) Experiments is the way forward
  16. M25 👍 1 Answered D
    Went with D

Reader comments & corrections

No reader corrections yet — spotted an error or have a note? Share it below.

Log in to comment, report an error, or add a note about this question.

How to Minimize Development Time for Weekly TensorFlow Retraining in Vertex AI?

S-grade

How to Handle Categorical Features in BigQuery ML for Fast Deployment?

S-grade

How to preprocess BigQuery data for TensorFlow?

S-grade

How to Track Model and Prediction Lineage in Vertex AI?

S-grade

Practice All PMLE Questions

Access 65 questions with complete answers and detailed explanations.

View Full PMLE Practice Test →

← Back to PMLE Study Guide