Skip to content

Uncertainty Estimation

Alviss AI is built on a Bayesian framework, providing robust uncertainty estimation across all results within the platform. Understanding and leveraging this feature can greatly enhance the decision-making process by providing insights into the confidence of predictions and results.

How Uncertainty is Quantified

  • Bayesian Framework: Alviss AI employs a Bayesian framework, which allows for comprehensive uncertainty estimation in all analyses and results.
  • Quantiles in Raw Data: In the raw data, uncertainty is quantified using quantiles. This method provides a clear and concise measure of variability and confidence intervals for predictions.

Quantiles Explained

Quantiles are a statistical tool used to understand the distribution of data by dividing it into equal-sized, contiguous intervals. Here’s how they work:

  • Definition: A quantile is a value that divides a dataset into equal-sized subsets. For example, the median is a quantile that divides the data into two equal halves.
  • Common Quantiles:
    • Median (50th percentile): The middle value of the data.
    • Quartiles: Divides the data into four equal parts (25th, 50th, 75th percentiles).
  • Usage in Alviss AI: Quantiles in Alviss AI are used to provide a range within which the results are expected to fall, giving a clear picture of the uncertainty and variability of the data.

Accessing Uncertainty Information

  • Data Exports: Uncertainty estimates are included in all data exports from Alviss AI. This ensures that any data taken out of the platform retains its confidence measures, providing consistent insights even outside the application.
  • API Responses: When interacting with Alviss AI via the API, uncertainty information is always included in the responses. This integration allows for seamless utilization of uncertainty data in automated workflows and external applications.

User Interface Experience

Uncertainty estimation is also integrated into the user interface (UI) of Alviss AI, enhancing the visual analysis experience.

  • Dashboard Filters: Uncertainty information can be accessed under the “Filter” function in the dashboards. This allows users to refine their views and better understand the range of possible outcomes.
  • Plots and Visuals: The representation of uncertainty varies depending on the plot being shown. Different types of visualizations might display uncertainty in various ways, but it is consistently available across the platform to aid in comprehensive analysis.

Conclusion

Leveraging the uncertainty estimation features in Alviss AI can provide deeper insights and more confident decision-making. Whether through direct data exports, API interactions, or the intuitive UI, uncertainty measures are readily accessible to enhance your analytical capabilities.