Grid Search vs. Random Search vs. Bayesian Optimization
Better methods for hyperparameter tuning.
The two most common approaches for hyperparameter tuning are:
Grid search
Random search
The visual below depicts how they work:
But they have many limitations.
For instance:
Grid search performs an exhaustive search over all combinations. This is computationally expensive.
Grid search and random search are restricted to the specified hyperparameter range. Yet, the ideal hyperparameter may exist outside that range.
They can ONLY perform discrete searches, even if the hyperparameter is continuous.
To this end, Bayesian Optimization is a highly underappreciated yet immensely powerful approach for tuning hyperparameters.
We covered this in detail with implementation here: Bayesian Optimization for Hyperparameter Tuning.
It uses Bayesian statistics to estimate the distribution of the best hyperparameters.
Here’s how it differs from Grid search and Random Search:
Both Grid search and Random Search evaluate every hyperparameter configuration independently. Thus, they iteratively explore all hyperparameter configurations to find the most optimal one.
However, Bayesian Optimization takes informed steps based on the results of the previous hyperparameter configurations.
This lets it confidently discard non-optimal configurations. Consequently, the model converges to an optimal set of hyperparameters much faster.
The efficacy of Bayesian Optimization is evident from the image below.
Bayesian optimization leads the model to the same F1 score but:
it takes 7x fewer iterations
it executes 5x faster
it reaches the optimal configuration earlier
Typically, Bayesian Optimization is only advised for intermediate/large models since small models can be trained in quick time with grid search and random search. In other words, the gain in speed-up isn’t significant for small models.
If you are curious about Bayesian Optimization, I once published a full deep dive on Bayesian optimization, which you can read here: Bayesian Optimization for Hyperparameter Tuning.
Why care?
The idea behind Bayesian optimization appeared to be extremely compelling to me when I first learned it.
Learning about this optimized hyperparameter tuning and utilizing it has been extremely helpful to me in building large ML models quickly.
Thus, learning about Bayesian optimization will be immensely valuable if you envision doing the same.
Assuming you have never had any experience with Bayesian optimation before, the article covers:
What is the motivation for Bayesian optimization?
How does Bayesian optimization work and its intuition?
Results from the research paper that proposed Bayesian optimization for hyperparameter tuning.
A hands-on Bayesian optimization experiment.
Comparing Bayesian optimization with grid search and random search.
Best practices for using Bayesian optimization.
👉 Interested folks can read it here: Bayesian Optimization for Hyperparameter Tuning.
Are you overwhelmed with the amount of information in ML/DS?
Every week, I publish no-fluff deep dives on topics that truly matter to your skills for ML/DS roles.
For instance:
A Crash Course on Graph Neural Networks (Implementation Included)
Conformal Predictions: Build Confidence in Your ML Model’s Predictions
Quantization: Optimize ML Models to Run Them on Tiny Hardware
5 Must-Know Ways to Test ML Models in Production (Implementation Included)
And many many more.
Join below to unlock all full articles:
SPONSOR US
Get your product in front of 85,000 data scientists and other tech professionals.
Our newsletter puts your products and services directly in front of an audience that matters — thousands of leaders, senior data scientists, machine learning engineers, data analysts, etc., who have influence over significant tech decisions and big purchases.
To ensure your product reaches this influential audience, reserve your space here or reply to this email to ensure your product reaches this influential audience.
Thank you for this, if I had known this earlier I would have used it for my msc data science project I’m doing right now and it would have saved me lots of time.