:notoc: .. module:: twinlab **************************************** twinLab - Probabilistic ML for Engineers **************************************** twinLab is a tool for augmenting your engineering workflows with Probabilistic Machine Learning. It enables you to quickly and easily build real-time emulators of your simulations, experimental set-ups, or sensor networks. Then you can make predictions, get recommendations, perform optimisations, and calibrate physics parameters from data. twinLab comes with built-in uncertainty quantification (UQ), which means that even with sparse or noisy data, you can maximise your understanding of the design space and surrogate model with confidence. For help, please email: `twinlab@digilab.co.uk `__ .. image:: _static/images/twinlab_light.png :align: center :width: 100% .. grid:: 1 2 2 2 :gutter: 4 :padding: 2 2 0 0 :class-container: sd-text-center .. grid-item-card:: Python Interface :link: python/index.html :class-card: intro-card :shadow: md Explore the reference guide, examples, and tutorials for the Python Interface, enabling easy workflow integration. +++ .. button-ref:: python :ref-type: ref :click-parent: :color: secondary :expand: Start solving → .. grid-item-card:: Speak to an expert :link: https://www.digilab.co.uk/contact :class-card: intro-card :shadow: md Our Solution Engineers are here to provide technical support and help you maximise the value of twinLab. +++ .. raw:: html Get in touch → Who is twinLab helping? ~~~~~~~~~~~~~~~~~~~~~~~ twinLab is solving engineering challenges for industry leaders. Here are some of them: .. image:: _static/images/twinlab_solving.png :align: center :width: 100% How does twinLab work? ~~~~~~~~~~~~~~~~~~~~~~ Under the hood, twinLab uses Gaussian Process emulators to analyse the often computationally expensive models that technical experts in industry rely on to make informed decisions about their engineering questions. The best way to use twinLab is via the crafted Python Interface. Why real-time emulators? ~~~~~~~~~~~~~~~~~~~~~~~~ Simulations, experiments, and sensors cost time and money to initialise and run. The necessary trade-offs result in poorly understood design spaces or sub-optimal parameter selection. twinLab enables you to build quick, cheap emulators that are simple to explore in real-time. Probabilistic ML means that you can understand model confidence, even if you start with sparse or noisy data. We also enable you to detach your emulator and embed it into systems to create a fully-responsive digital twin. Example case study ~~~~~~~~~~~~~~~~~~ The UK Atomic Energy Authority achieved a 10,000x speed-up over conventional parameter studies (i.e. days to seconds), with less than 2% model uncertainty. The output was an interoperable, real-time prediction tool. .. image:: _static/images/ukaea_case_study.png :align: center :width: 100% What can you do with twinLab? ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Predict**: Make real-time predictions with confidence, even with limited, noisy or uncertain datasets. Leverage Machine Learning to augment your expert opinions. - **Reduce Uncertainty**: Augment your sampling or test strategy with our `Recommend`` functionality. That means fewer simulations or experiments, with increased confidence. - **Calibrate**: Find best-fit values across all possible inputs and conditions, and maintain your understanding of your system even with only partial data. - **Optimise**: Powerful Bayesian optimisation techniques enable you to find the best answer in the shortest possible time, with assurance. Getting started ~~~~~~~~~~~~~~~ twinLab is proprietary software that requires an API key to access. We offer a range of deployment and licensing options for both small and large organisations. twinLab is cloud-hosted by default - contact us if you require on-premises deployment. To arrange a free trial, or to learn more about how probabilistic ML can augment your workflow, please email: `twinlab@digilab.co.uk `_ or fill in the contact form `here `_. **Useful Contacts**: `Q&A support `__ | `Mailing list <./changelog.rst>`__ .. toctree:: :maxdepth: 3 :hidden: :titlesonly: python/index RESTful/index changelog learn