The ML Hero’s Secret Weapon: AutoML

This is the AutoML presentation I used during GDG DevFest Turkey 2020.

Transcript

1. The ML Hero’s Secret
   Weapon: AutoML
   Daron Yöndem
   http://daron.me
   @daronyondem
   

   View Slide

2. What is it?
   • Automated Machine Learning (AutoML) is the process of
   automating end-to-end process of applying Machine Learning
   (ML) to create, develop and deploy predictive models so that
   any enterprise benefits from data.
   • Low-code / No-Code Experience
   • Keeps searching for the algorithm and hypermeters based on
   the metrics you define.
   

   View Slide

3. The Process
   SQL DB
   Cosmos DB
   Datawarehouse
   Data lake
   Blob storage
   …
   Prepare Data Build & Train Deploy
   

   View Slide

4. Sample Problem
   How much is this car worth?
   

   View Slide

5. Model creation takes time… a lot o time…
   Which algorithm? Which parameters?
   Which features?
   Mileage
   Condition
   Car brand
   Year of make
   Regulations
   …
   Gradient Boosted
   Nearest Neighbors
   SVM
   Bayesian Regression
   LGBM
   …
   Nearest Neighbors
   50%
   Model
   Iterate
   30%
   Gradient Boosted
   Mileage
   Car brand
   Year of make
   Car brand
   Year of make
   Condition
   Parameter 1
   Parameter 2
   Parameter 3
   Parameter 4
   …
   

   View Slide

6. The Complexity
   

   View Slide

7. Model Selection & Hyperparameter Tuning
   Dataset
   Training
   Algorithm 1
   Hyperparameter
   Values – config 1
   Model 1
   Hyperparameter
   Values – config 2
   Model 2
   Hyperparameter
   Values – config 3
   Model 3
   Model Training
   Infrastructure
   Training
   Algorithm 2
   Hyperparameter
   Values – config 4
   Model 4
   Repetitive & Manual
   

   View Slide

8. Introducing Automated Machine Learning
   Dataset
   Optimization
   Metric
   Constraints
   (Time/Cost)
   ML Model
   Automated ML
   Accessible & Faster
   

   View Slide

9. Data Scientist in a Box
   Demo
   

   View Slide

10. Thanks
    http://daron.me | @daronyondem
    Download slides here;
    http://daron.me/decks
    

    View Slide