Use SigOpt for any Model
Our mission is to accelerate and amplify the impact of modelers everywhere. With this in mind, we designed our solution to be fully agnostic as to the model type so it can be fully used and useful for any modeler. This, of course, is easier to show rather than tell.
The easiest way to do this would be to provide use cases for how our customers are using the product. But for us, the privacy, security and safety of our customer work comes first. So instead, we spend time building our own use cases that show you the potential for our solution for a variety of tasks. This is not intended to be comprehensive by any means – it is just a small sampling of what can be done using our product.
It seems like breakthroughs for novel applications of deep learning happen daily. Configuring hyperparameters can be the difference between success and failure of these models at any given task. SigOpt makes it easy to supercharge this hyperparameter optimization process. Here are a few examples:
Whether you need marginal performance gains or simply require certainty that you have configured your hyperparameters, SigOpt is a useful solution for any traditional machine learning model. Using SigOpt as early as possible in the model development process also ensures that you can iterate quickly through experiments to uncover the best models for your tasks. Here are a few examples:
Simulations, Reinforcement Learning & Other
Simulations are critical to the research process for many industries, including algorithmic trading, self-driving cars, materials, physical products, energy and manufacturing, among others. In some cases, simulations are the model being developed. In others, they are part of the process for evaluating the performance of models that are being developed to solve a given problem. In either case, SigOpt can be used to tune the parameters. Here are a few examples: