Open TTK95 opened 1 year ago
Thanks for suggesting @TTK95,
Notice that ClearML plots already provide a scaling control (see zoom control) to let you focus on any area of your plot that is most interesting.
Is this what you were looking for?
Thank you, @ainoam, for your response. However, the feature I am proposing is slightly different from the zoom control or scaling control that you have mentioned.
While the zoom control allows users to focus on a specific area of the plot, it does not address the issue of outliers skewing the scaling of the plot. My proposed feature would allow users to ignore outliers when scaling their plots, enabling them to create more accurate visualizations of their data.
To clarify further, this feature would not involve smoothing the data or adjusting the axis scales manually. Instead, it would be an autoscale feature that automatically adjusts the scaling of the plot to exclude outliers, making it easier to visualize the majority of the data points.
bump on this request, this is a standard feature on Tensorboard and would be really useful here!
Thanks for the nudge @JChunX 🙂 We're looking out for the best opportunity add this.
Proposal Summary
I propose the addition of a new feature to charting tools that would allow users to ignore outliers when scaling charts. This feature would be useful for cases where outliers can skew the visualization of data, making it difficult to see the majority of the data points.
Motivation
In many data sets, there can be extreme values that fall outside the typical range of the data. These outliers can have a disproportionate impact on the scaling of a chart, making it difficult to accurately visualize the majority of the data points.
By allowing users to ignore outliers when scaling charts, this feature would enable them to create more accurate and useful visualizations of their data. This would be especially beneficial for data scientists, analysts, and researchers who rely on data visualization to gain insights and communicate their findings to others.