VIP20
本课程属于：005-Artificial Intelligence and Machine Learning Engineering English Technical Courses · 共 20 门相关课程，组合学习更完整。
查看所属VIP全部课程 →
← 返回所属目录 咨询 / 购买

Automated Model Selection and Tuning for Industrial AI

30 Chapters
View Course Samples Online
01
Industrial AI landscape
defining automated model selection and tuning, key challenges in industrial settings, and the end-to-end pipeline from data acquisition to deployment.
Industriallandscape
02
Industrial data characteristics
sensor time series, high-dimensional process data, class imbalance, concept drift, and data quality issues that impact model selection and tuning.
Industrialcharacteristics
03
Problem formulation for industrial tasks
classification, regression, anomaly detection, remaining useful life estimation, and their mapping to model selection objectives.
formulationindustrial
04
Model selection fundamentals
bias-variance tradeoff, cross-validation strategies for temporal and grouped industrial data, and performance metrics aligned with business KPIs.
selectionbias-variance
05
Hyperparameter tuning basics
grid search, random search, and their computational limits in large-scale industrial model spaces.
Hyperparametercomputational
06
Bayesian optimization for hyperparameter tuning
Gaussian processes, acquisition functions, and application to expensive industrial model evaluations.
Bayesianoptimization
07
Advanced hyperparameter optimization
Hyperband, BOHB, population-based training, and multi-fidelity methods for resource-constrained industrial environments.
Advancedhyperparameter
08
Automated machine learning (AutoML) frameworks
overview of TPOT, Auto-sklearn, H2O, and their suitability for industrial tabular and time-series data.
AutomatedAutoML
09
Neural architecture search (NAS)
reinforcement learning, evolutionary, and differentiable methods, with industrial constraints on latency and memory.
architectureNAS
10
Feature engineering automation
automated feature extraction, selection, and construction for industrial sensor and log data.
automationautomated
11
Pipeline automation with scikit-learn
building composite estimators, using Pipeline and ColumnTransformer, and integrating hyperparameter search.
Pipelineautomation
12
Industrial case study
automated model selection for predictive maintenance using vibration and acoustic data, including tuning of window sizes and model hyperparameters.
Industrialautomated
13
Industrial case study
tuning anomaly detection models for manufacturing quality control, covering isolation forests, autoencoders, and threshold optimization.
Industrialdetection
14
Industrial case study
automated selection and tuning of regression models for soft sensors in chemical processes, addressing drift and label noise.
Industrialautomated
15
Handling concept drift in automated tuning
online hyperparameter adaptation, sliding-window validation, and drift detection triggers for re-tuning.
Handlingautomated
16
Multi-objective optimization in industrial AI
balancing accuracy, inference latency, model size, and interpretability using Pareto fronts and NSGA-II.
Multi-objectiveoptimization
17
Resource-aware tuning
distributed and parallel hyperparameter search, early stopping, and scheduling on edge and cloud infrastructure.
Resource-awaredistributed
18
Transfer learning and meta-learning for model selection
leveraging similar industrial tasks to warm-start hyperparameter optimization.
Transfermeta-learning
19
Automated tuning for deep learning in industrial vision
learning rate schedules, batch size, augmentation policies, and architecture choices.
Automatedindustrial
20
Automated tuning for time-series forecasting in industrial energy management
window sizes, model types, and exogenous variable selection.
Automatedtime-series
21
Model selection under label scarcity
semi-supervised and active learning strategies combined with hyperparameter tuning for industrial defect detection.
selectionscarcity
22
Interpretability-aware model selection
incorporating SHAP, LIME, and feature importance into the tuning objective for regulated industries.
Interpretability-awareselection
23
Reproducibility and versioning in automated tuning
experiment tracking, model registries, and CI/CD for industrial ML pipelines.
Reproducibilityversioning
24
Deployment considerations
model compression, quantization, and hardware-specific tuning for edge industrial devices.
Deploymentconsiderations
25
Monitoring and maintenance of tuned models
performance degradation detection, automated re-tuning triggers, and feedback loops.
Monitoringmaintenance
26
Benchmarking and evaluation
designing fair comparisons of AutoML methods on industrial datasets, including statistical significance and robustness.
Benchmarkingevaluation
27
Scalable infrastructure for automated tuning
Kubernetes, Ray Tune, Optuna, and cloud-based hyperparameter optimization services.
Scalableinfrastructure
28
Industrial case study
end-to-end automated model selection and tuning for a steel defect classification system, from data ingestion to edge deployment.
Industrialend-to-end
29
Industrial case study
automated tuning of reinforcement learning agents for process control, addressing safety constraints and sample efficiency.
Industrialautomated
30
Future directions and responsible AI
emerging trends in automated model selection, ethical considerations, and integration with digital twins and Industry 4.0.
directionsresponsible