Published Oct 20, 2020

When data leakage turns into a flood of trouble

Explore the ramifications of data leakage in machine learning with insights from Rajiv Shah on enhancing model robustness through partitioning techniques, skepticism in model performance, and the use of activation maps to safeguard against skewed outcomes, ensuring reliable and effective deployment.

Episode Highlights

Topics covered

Episode Highlights

Related Episodes

Building a career in Data Science
Answers 383 questions
Data science for intuitive user experiences
Answers 383 questions
What exactly is "data science" these days?
Answers 383 questions
The ins and outs of open source for AI
Answers 383 questions
Machine learning at small organizations
Answers 383 questions
Data for All
Answers 383 questions
When AI goes wrong
Answers 383 questions
Explaining AI explainability
Answers 383 questions
AI code that facilitates good science
Answers 383 questions
Open source data labeling tools
Answers 383 questions
Towards stability and robustness
Answers 383 questions
Escaping the "dark ages" of AI infrastructure
Answers 383 questions
Using edge models to find sensitive data
Answers 383 questions
The influence of open source on AI development
Answers 383 questions
Testing ML systems
Answers 383 questions

Dexa/Practical AI

When data leakage turns into a flood of trouble

Topics covered

Popular Clips

Data Leakage Concerns

AI Insights

Target Leakage Challenges

AI Insights

Data Leakage Insights

Data Leakage Pitfalls

Data Partitioning Insights

Practical AI Insights

AI Conversations

Data Partitioning Best Practices

Data Science Skills

Data Leakage Awareness

Data Science Pitfalls

Data Partitioning Insights

Episode Highlights

Partitioning Data Correctly

Effective Data Science Practices

Understanding Data Leakage

Related Episodes

Building a career in Data Science

Data science for intuitive user experiences

What exactly is "data science" these days?

The ins and outs of open source for AI

Machine learning at small organizations

Data for All

When AI goes wrong

Explaining AI explainability

AI code that facilitates good science

Open source data labeling tools

Towards stability and robustness

Escaping the "dark ages" of AI infrastructure

Using edge models to find sensitive data

The influence of open source on AI development

Testing ML systems

When data leakage turns into a flood of trouble

Topics covered

Popular Clips

Episode Highlights

Partitioning Data CorrectlyRajiv Shah explores effective data partitioning methods and the use of nested cross-validation to enhance machine learning model robustness. He emphasizes the importance of skepticism in model performance to prevent overfitting and ensure successful deployment.

Partitioning Data Correctly

Effective Data Science PracticesRajiv Shah discusses the critical issue of data leakage in machine learning models and its impact on production. He shares insights on using activation maps and image embedding to detect and prevent leakage, ensuring robust model performance.

Effective Data Science Practices

Understanding Data LeakageRajiv Shah sheds light on the pervasive issue of data leakage in machine learning, emphasizing its potential to skew model performance. He shares insights into identifying and preventing leakage, ensuring more reliable and accurate predictive models.

Understanding Data Leakage

Related Episodes