Building and Optimizing AI Models

本课程是 Transformers Unleashed: Master the Architecture of Modern AI 专业证书的一部分

位教师：Professionals from the Industry

访问权限由 Coursera Learning Team 提供

8个模块

深入了解一个主题并学习基础知识。

中级等级

推荐体验

8 小时完成

灵活的计划

自行安排学习进度

8个模块

深入了解一个主题并学习基础知识。

中级等级

推荐体验

8 小时完成

灵活的计划

自行安排学习进度

您将学到什么

Train and evaluate predictive machine learning models using supervised and unsupervised algorithms
Design custom neural network architectures for AI applications
Optimize deep learning models using transfer learning and performance tuning
Benchmark AI algorithms to evaluate efficiency, accuracy, and computational cost

您将获得的技能

您将学习的工具

要了解的详细信息

可分享的证书

添加到您的领英档案

作业

20 任务¹

AI 评分请参见免责声明

授课语言：英语（English）

了解顶级公司的员工如何掌握热门技能

了解关于 Coursera for Business 的更多信息

Petrobras, TATA, Danone, Capgemini, P&G 和 L'Oreal 的徽标

积累 Machine Learning 领域的专业知识

本课程是 Transformers Unleashed: Master the Architecture of Modern AI 专业证书专项课程的一部分

在注册此课程时，您还会同时注册此专业证书。

向行业专家学习新概念
获得对主题或工具的基础理解
通过实践项目培养工作相关技能
通过 Coursera 获得可共享的职业证书

该课程共有8个模块

Building and Optimizing AI Models introduces the foundational engineering practices required to design, train, and optimize machine learning models for modern AI systems. In this course, you will explore statistical machine learning methods, neural network architectures, and deep learning optimization techniques used to develop high-performing predictive models.

You will begin by applying supervised and unsupervised algorithms to train and evaluate predictive models. Next, you will design custom neural network architectures and experiment with different layer configurations to improve model accuracy and efficiency. The course also introduces transfer learning and deep learning optimization strategies that help adapt pretrained models to domain-specific tasks. Finally, you will analyze algorithm performance and benchmark model implementations to understand trade-offs between accuracy, latency, and computational cost. By the end of this course, you will be able to design neural networks, optimize deep learning workflows, and evaluate model performance using industry-standard metrics. Tools and technologies covered include Python, TensorFlow, neural network frameworks, and model performance benchmarking techniques.

You will apply supervised and unsupervised algorithms to train predictive models using structured datasets. You will implement cross-validation techniques to validate model reliability and interpret results to ensure robust performance.

涵盖的内容

3个视频1篇阅读材料2个作业

3个视频总计17分钟

Welcome and What You’ll Learn4分钟
Supervised vs. Unsupervised Modeling: When to Use Each5分钟
Walkthrough: Training Logistic Regression and K-Means in scikit-learn8分钟

1篇阅读材料总计10分钟

How Cross-Validation Improves Model Reliability10分钟

2个作业总计22分钟

Hands-On Activity: Train Two Models and Run 5-Fold CV15分钟
Practice Quiz: Model Fit Check7分钟

You will evaluate model performance using accuracy and F1 metrics, identify weaknesses, and refine features systematically. You will iterate on feature engineering decisions to meet defined performance targets

涵盖的内容

3个视频1篇阅读材料3个作业

3个视频总计15分钟

Why Metrics Drive Better Modeling4分钟
Interpreting Accuracy, Precision, Recall, and F17分钟
Demo: Interaction Features Improve F14分钟

1篇阅读材料总计10分钟

Feature Engineering Fundamentals: Transform, Combine, Improve10分钟

3个作业总计42分钟

Graded Quiz: Build, Validate, and Improve a Predictive Model20分钟
Hands-On Activity: Improve a Model’s F1 Score with New Features15分钟
Practice Quiz: Fix the Model7分钟

You will analyze candidate neural network topologies such as CNNs, RNNs, and Transformers. You will evaluate task requirements, data characteristics, and compute constraints to select the most appropriate architecture.

涵盖的内容

3个视频1篇阅读材料2个作业

3个视频总计8分钟

Welcome and Why Architecture Choices Matter2分钟
Comparing Neural Network Topologies3分钟
How to Evaluate Architecture Fit in Practice3分钟

1篇阅读材料总计10分钟

Understanding Task, Data, and Compute Constraints10分钟

2个作业总计22分钟

Hands-on Activity: Choose the Best Architecture Under Real Constraints15分钟
Practice Quiz: Architecture Selection Mini-Review7分钟

You will create custom neural-network architectures by composing layers, activations, and regularization techniques. You will test architectural decisions to improve generalization and training stability.

涵盖的内容

3个视频1篇阅读材料3个作业

3个视频总计9分钟

Why Build Custom Architectures2分钟
Layers, Activations, and Regularization2分钟
Screencast: Constructing a Custom Model in PyTorch5分钟

1篇阅读材料总计10分钟

Designing a Custom Network Step by Step10分钟

3个作业总计42分钟

Graded Assessment: Custom Neural Network Architechture Evaluation20分钟
Hands-on Activity: Build Your Own Network Architecture15分钟
Practice Quiz: Improve a Baseline Model With Regularization7分钟

You will apply transfer-learning workflows to fine-tune pretrained models on domain-specific datasets. You will experiment with freezing and unfreezing layers to improve model adaptation.

涵盖的内容

3个视频1篇阅读材料2个作业

You will evaluate deep model configurations by comparing accuracy, latency, and memory usage. You will balance performance and efficiency to determine the most suitable production-ready configuration.

涵盖的内容

3个视频1篇阅读材料3个作业

3个视频总计17分钟

Accuracy vs. Efficiency: The Real Trade-Offs6分钟
Hyperparameter Sweeps: Comparing Configurations Fairly (Optuna Example)5分钟
Quantization as a Configuration Choice: Speed vs. Accuracy (TensorRT Example)6分钟

1篇阅读材料总计10分钟

Practical Model Training Tips for Reliable Machine Learning Performance10分钟

3个作业总计40分钟

Graded Assessment: Model Optimization Decision Challenge20分钟
Hands-On Activity: Run a Mini Optimization Comparison15分钟
Practice Quiz: Evaluating Model Performance Trade-Offs5分钟

You will analyze the computational complexity of algorithms and evaluate how data structures affect performance. You will select optimal approaches based on scalability and workload demands.

涵盖的内容

3个视频1篇阅读材料2个作业

3个视频总计15分钟

Welcome and Why Speed Matters in Real AI Systems4分钟
Understanding Complexity: From Big-O to Practical Speed5分钟
Hidden Costs: Constants, Cache Effects, and Real-World Slowdowns6分钟

1篇阅读材料总计10分钟

Data Structures That Scale: Trees, Hash Maps, and Heaps10分钟

2个作业总计20分钟

Hands-On Activity: Complexity Match-Up: Predict the Faster Method10分钟
Practice Quiz: Test Your Complexity and Data Structure Skills10分钟

You will create prototype algorithms and design structured benchmarks to measure latency, throughput, and memory usage. You will interpret benchmark results to evaluate performance trade-offs and justify implementation decisions.