![]() These methods are able to learn from incomplete data, and then Value prediction methods based on deep learning algorithms. To this end, Azua provides state-of-the-art missing Make accurate estimates regarding the missing information. In many real-life scenarios, we will need to make decisions under incomplete information. Core functionalitiesĪzua has there core functionalities as shown below depends on the type of decision maksing tasks. In NeurIPS: Workshop on Deep Generative Models and Applications (2021) Resourcesįor quick introduction to our work regarding best next question, checkout our NeurIPS 2020 tutorial, from 2:17:11.įor a more in-depth technical introduction of deep genertive model for missing value prediction and best next question, checkout our ICML 2020 tutorial 1. ![]() Accurate Imputation and Efficient Data Acquisition with , (Transformer PVAE, Transformer encoder PVAE, Rupert) Sarah Lewis, Tatiana Matejovicova, Angus Lamb, Yordan Zaykov, In Advances in Neural Information Processing Systems 34 (2021) Identifiable Generative Models for Missing Not at Random Data Imputation. In GReS: Workshop on Graph Neural Networks for Recommendation and Search, 2021 CORGI: Content-Rich Graph Neural Networks with Attention. , (CORGI:) Jooyeon Kim, Angus Lamb, Simon Woodhead, Simon Pyton Jones, Cheng Zhang, and Miltiadis Allamanis. "Results and Insights from Diagnostic Questions: The NeurIPS 2020 Education Challenge." arXiv preprint arXiv:2104.04034 (2021). , (Eedi dataset) Zichao Wang, Angus Lamb, Evgeny Saveliev, Pashmina Cameron, Yordan Zaykov, Jose Miguel Hernandez-Lobato, Richard E. "VAEM: a Deep Generative Model for Heterogeneous Mixed Type Data." Advances in Neural Information Processing Systems 33 (2020). , (VAEM) Chao Ma, Sebastian Tschiatschek, Richard Turner, José Miguel Hernández-Lobato, and Cheng Zhang. "EDDI: Efficient Dynamic Discovery of High-Value Information with Partial VAE." In International Conference on Machine Learning, pp. , (PVAE and information acquisition) Chao Ma, Sebastian Tschiatschek, Konstantina Palla, Jose Miguel Hernandez-Lobato, Sebastian Nowozin, and Cheng Zhang. If you have used the models in our code base, please consider to cite the corresponding paper: For commercial applications, please reach out to us at if you are interested in using our technology as a service. ![]() Our technology has enabled personalized decision-making in real-world systems, combining multiple advanced research methodologies in simple APIs suitableįor research development in the research community, and commercial use by data scientists and developers. We also provide the flexibility to use any core functionalities such as missing value prediction, best next question, etc, separately depending on the users' needs. With these decision-making goals, one can use our codebase in an end-to-end way for decision-making. Our technology for "best next question" decisions is driven by state-of-the-art algorithms for Bayesian experimental design and active learning. Thus, the first part of project Azua focuses on enabling machine learning solutions to gather personalized information, allowing the machine to know the "best next question" and make a final judgment efficiently. Humans are very efficient at gathering information and drawing the correct conclusion, while most deep learning methods require significant amounts of training data. In daily life, one type of decision we make relates to information gathering for "get to know" decisions for example, a medical doctor takes a medical test to decide the correct diagnosis for a patient. Our conceptual framework is to divide decisions into two types: "best next question" and "best next action". Project Azua aims to develop machine learning solutions for efficient decision making that demonstrate human expert-level performance across all domains. Humans make tens of thousands of decisions every day.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |