We’re proud to share after months of work and layers upon layers of engineering, we have officially deployed cloud optimization to parameter sensitivity test your strategies. The new systems allow you to test thousands of combinations of your strategy parameters to see areas of strength and weakness, to avoid strategies sensitive to narrowly defined parameters. The technology supports all asset classes, data resolutions, and styles of algorithms built-in QuantConnect, including optimizing machine learning.

Such technology is virtually unparalleled in the quantitative space. The QuantConnect cloud optimization cuts the guesswork and iteration involved with algo parameter testing. Manually selecting parameters and re-running backtests — previously, the only option to tune your algos — can often lead to the deployment of overfit strategies, which can incur significant losses if deployed live. Now, with cloud optimization, you can streamline your production time and run up to 24 different tests in parallel to generate heat maps of the parameter contributions to the overall result. 

Often the network topology can be an important contributing factor for neural network fit; resulting in machine learning engineers spending a lot of time exploring different topologies or resorting to deep learning techniques where the topology is unknown. With the new QuantConnect Optimizer you can explore a range of network designs. 

To see the results of your optimization, we compile the results into a sleek graphical user interface that provides insights into how the parameters you’ve selected impact your alpha. With this charting, you can easily compare and contrast the high and low points to best determine the most efficient parameters for your strategy.

While we encourage you to harness the computing power of the cloud, this robust tool is also available to run locally on LEAN. This open-sourced version allows you to run similar batch processing locally and plug the results into the UX of your choice.

We are proud to be pioneering the way for radically open-source quantitative finance and will continue to strive to open the doors of quantitative finance, delivering institutional-grade tools and resources to our community at a low cost.