Not a Magic - What to Expect from Machine Learning Projects

Explore the intricacies of implementing Machine Learning solutions in organizations through this comprehensive conference talk. Gain insights into defining possibilities, identifying starting points, navigating common pitfalls, managing expectations, and delivering tangible value within reasonable timeframes. Delve into crucial aspects of ML projects, including data mining, protection, cleaning, and domain understanding. Learn about the nature of data, progress sharing, and reporting techniques. Examine learning algorithms, training processes, feature engineering, and model evaluation methods for supervised learning. Discover the realities of data pipelines, monitoring strategies, and model deployment. Understand the importance of A/B testing and proof of concept in ML projects. Equip yourself with practical knowledge to successfully navigate the challenges of Machine Learning implementation in organizational settings.

Syllabus

Intro
Machine Learning Project
Data mining
Data protection
Data cleaning
Data domain
The nature of Data
Data – sharing the progress
Data reporting
Learning data
Learning algorithms
Training: Loss and Cost funtion
Training: Objective function and loss
Training: summary
Feature engineering
Evaluate Metric: Supervised learning
Evaluate model: high variance
Evaluate: compare with what
Evaluate model: same distirbution
Deep learning
Data Pipeline - reality
Monitoring
Reports
Model deployment
A/B testing
Proof of Concept (PC)
Modelling