Want to learn more about the technologies behind SigOpt? We’ve got you covered.

Introductory Materials

Get started with the basics behind SigOpt’s methods.

Common Problems in Hyperparameter Optimization

A discussion of common problems observed in hyperparameter optimization

Bayesian Optimization Primer

An introductory document discussing the key concepts behind Bayesian optimization.

Intuition Behind Gaussian Processes

A primer on Gaussian processes and how they power SigOpt’s optimization.

Intuition Behind Covariance Kernels

Learn about covariance kernels and how they apply to Gaussian processes.

Machine Learning

Discussions of applying SigOpt's methods to Machine Learning and Hyperparameter Optimization.

Deep Learning Hyperparameter Optimization with Competing Objectives

Optimizing deep learning models with multiple objectives

Using Bayesian Optimization for Reinforcement Learning

Using SigOpt and reinforcement learning to play games with OpenAI

Automatically Tuning Text Classifiers

A short example using SigOpt and scikit-learn to build and tune a text sentiment classifier.

Unsupervised Learning with Bayesian Optimization

An example using SigOpt and xgboost to build an optical character recognition model using unsupervised techniques.

Building Convolutional Neural Networks with Bayesian Optimization

An example using SigOpt and TensorFlow to build a CNN for optical character recognition.

Deep Neural Network Optimization with SigOpt and Nervana Cloud

Optimizing deep neural nets with SigOpt and Nervana Cloud.

Bayesian Optimization for Collaborative Filtering

An example using SigOpt and MLlib to optimize parameters of the alternating least squares algorithm.

Tuning Machine Learning Models

A comparison of different hyperparameter optimization methods.

Active Preference Learning for Personalized Portfolio Construction

Using financial asset management as a case study, discuss using preference learning to construct a diverse set of optimal solutions

Parameter Optimization

Discussions of problem types where SigOpt can be used.

Multicriteria Optimization

A short example using SigOpt to tune a system with multiple independent objectives.

Preemptive Termination of Suggestions during Sequential Kriging Optimization

Applications of Bayesian optimization

Interactive Preference Learning of Utility Functions for Multi-Objective Optimization

A system for more efficient multicriteria optimization

Making a Better Airplane using SigOpt and Rescale

An example of deploying and scaling parameter optimization methods.

Winning on Wall Street: Tuning Trading Models with Bayesian Optimization

Using SigOpt to tune a financial trading model.

Using Model Tuning to Beat Vegas

Using SigOpt to tune a model for predicting basketball scores.

Robust Bayesian Optimization with Student-t Likelihood

A discussion of Robust Bayesian Optimization, which can be more resistant to outlier data

Practical Bayesian optimization in the presence of outliers

Detecting and optimizing in the face of outliers

Filtering Outliers in Bayesian Optimization

A Bayesian optimization algorithm to autonomously manage situations where outlier data may be encountered

Policy Search using Robust Bayesian Optimization

Applying Robust Bayesian Optimization to robotics applications

Comparisons to Other Methods

An overview of how we rigourously compare SigOpt to other optimization methods.

A Stratified Analysis of Bayesian Optimization Methods

An article outlining an aggregative performance comparison strategy for Bayesian optimization methods.

Evaluation System for a Bayesian Optimization Service

An article presenting an overview of SigOpt’s evaluation framework system used to benchmark our optimization engine.

A Strategy for Ranking Optimization Methods using Multiple Criteria

A more abstract framework for comparing optimization methods.

Gaussian Processes

Approximation of Data

Techniques on making predictions from incomplete data.

Likelihood for Gaussian Processes

Using likelihood to choose the most appropriate Gaussian processes.

Profile Likelihood vs. Kriging Variance

A comparison of different Gaussian Process strategies

Expected Improvement vs. Knowledge Gradient

A comparison of different acquisition functions.

Covariance Kernels for Avoiding Boundaries

Learn how SigOpt adapts to problems where the boundaries should be avoided.