Evaluation of Cloud Hosting Frameworks for Machine...
summary
Builds regression‑based anomaly detectors for alternator coil temperatures using pandas, scikit-learn, Keras and ADTK, serializes them as pickle files and wraps pre/post processing into a predictor class. Compares deployment on GCP, Azure and AWS by measuring wrapper code, runtime compatibility (Python 3.5 vs 3.6, sklearn 0.21→0.22 breaks), and ops friction: GCP gives high‑level ML API, AWS pushes containers and IAM complexity; models deploy successfully to two providers but require version and packaging workarounds.
AI-written summary · about
chapters (10)
- 0:00Cloud hosting for ML monitoring
- 5:00Data science to production challenges
- 6:24ML for equipment monitoring
- 9:14Predicting alternator anomalies
- 11:44Cloud services for model deployment
- 15:21Deployment scenarios compared
- 24:15Implementation and docs gaps
- 28:51Customizing models
- 31:00Deploying custom packages
- 33:06Key takeaways
related talks
40mCloud-based Machine Learning Offerings - Beginner's Perspective
35mDoing Data Science on Cloud
41mBuilding Serverless Machine Learning Models in the Cloud
12mAI-Powered Predictive Analytics for Multi-Cloud Management
44mPragmatic Machine Learning in the Cloud
17mProductionizing Data Science Workloads
22mOperationalising Data Science on Cloud Foundry
34mEmpowering AWS DevOps with Python and Machine Learning