Major changes on the Experiment data structure
#14
Labels
new feature
New feature request
Comments
wookayin
added a commit
that referenced
this issue
Nov 27, 2022
- df: a MultiIndex-ed DataFrame w.r.t config keys - config_keys: The list of (index) keys to run config, e.g., hyperparameters to compare. - summary_columns: The list of additional columns to be included in the dataframe, obtained via `Hypothesis.summary`. Note: `ex.plot()` will now respect this information rather than drawing for all the columns.
wookayin
added a commit
that referenced
this issue
Nov 27, 2022
wookayin
added a commit
that referenced
this issue
Dec 13, 2022
These are shorthands for RunList.to_dataframe() followed by Experiment.from_dataframe(). See #14.
wookayin
added a commit
that referenced
this issue
Dec 13, 2022
With the dataframe structure enclosed in Experiment (#14), we can now have a more flexible pandas-like API (see df.query) to select a subset of hypotheses in an Experiment.
wookayin
added a commit
that referenced
this issue
Dec 19, 2022
`Experiment.with_config_keys(new_config_keys)` returns a new Experiment object with the same hypotheses but with then another list of config keys; useful for reordering the hyperparameters. See #14.
wookayin
added a commit
that referenced
this issue
Feb 14, 2023
A filter function fn: Hypothesis -> bool can be used to select a subset of rows in the Experiment. This generalizes the previous method Experiment.select(expr) where expr is a pandas-compatible query string expression. See #14
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Goal: Make
Experimenta hierarchically structured data structure (with a MultiIndex-ed DataFrame) of multiple hypotheses, providing a lot of handy APIs to query and select individual hypotheses and runs.The text was updated successfully, but these errors were encountered: