Remove ranking from all tables

Author: ElliottKasoar

docs/source/developer_guide/add_benchmarks.rst

@@ -262,10 +262,10 @@ For ``@build_table``, the value returned should be of the form:
         ...
     }
 
-This will generate a table with columns for each metric, as well as "MLIP", "Score",
-and "Rank" columns. Tooltips for each column header can also be set by the decorator,
-as well as the location to save the JSON file to be loaded when building the app,
-which typically would be placed in ``ml_peg/app/data/[category]/[benchmark_name]``.
+This will generate a table with columns for each metric, as well as "MLIP" and "Score"
+columns. Tooltips for each column header can also be set by the decorator, as well as
+the location to save the JSON file to be loaded when building the app, which typically
+would be placed in ``ml_peg/app/data/[category]/[benchmark_name]``.
 
 Every benchmark should have at least one of these tables, which includes
 the score for each metric, and allowing the table to calculate an overall score for the

ml_peg/analysis/utils/decorators.py

@@ -12,7 +12,7 @@
 import numpy as np
 import plotly.graph_objects as go
 
-from ml_peg.analysis.utils.utils import calc_ranks, calc_table_scores
+from ml_peg.analysis.utils.utils import calc_table_scores
 from ml_peg.app.utils.utils import Thresholds
 
 
@@ -274,8 +274,7 @@ def build_table(
     filename
         Filename to save table. Default is "table.json".
     metric_tooltips
-        Tooltips for table metric headers. Defaults are set for "MLIP", "Score", and
-        "Rank".
+        Tooltips for table metric headers. Defaults are set for "MLIP" and "Score".
     normalize
         Whether to apply normalisation when calculating the score. Default is True.
     normalizer
@@ -365,10 +364,7 @@ def build_table_wrapper(*args, **kwargs) -> dict[str, Any]:
                     | {"id": mlip},
                 )
 
-            summary_tooltips = {
-                "MLIP": "Name of the model",
-                "Rank": "Model rank based on score (lower is better)",
-            }
+            summary_tooltips = {"MLIP": "Name of the model"}
             if normalize:
                 summary_tooltips["Score"] = (
                     "Average of normalised metrics (higher is better)"
@@ -391,8 +387,7 @@ def build_table_wrapper(*args, **kwargs) -> dict[str, Any]:
             else:
                 metrics_data = calc_table_scores(metrics_data)
 
-            metrics_data = calc_ranks(metrics_data)
-            metrics_columns += ("Score", "Rank")
+            metrics_columns += ("Score",)
 
             metric_weights = weights if weights else {}
             for column in results:

ml_peg/analysis/utils/utils.py

@@ -8,7 +8,6 @@
 from matplotlib import cm
 from matplotlib.colors import Colormap
 import numpy as np
-from scipy.stats import rankdata
 from sklearn.metrics import mean_absolute_error, mean_squared_error
 from yaml import safe_load
 
@@ -145,7 +144,7 @@ def calc_metric_scores(
     for row in metrics_scores:
         for key, value in row.items():
             # Value may be ``None`` if missing for a benchmark
-            if key not in {"MLIP", "Score", "Rank", "id"} and value is not None:
+            if key not in {"MLIP", "Score", "id"} and value is not None:
                 if cleaned_thresholds is None or key not in cleaned_thresholds:
                     row[key] = value
                     continue
@@ -194,7 +193,7 @@ def calc_table_scores(
         weights_list = []
         for key, value in metrics_row.items():
             # Value may be ``None`` if missing for a benchmark
-            if key not in {"MLIP", "Score", "Rank", "id"} and value is not None:
+            if key not in {"MLIP", "Score", "id"} and value is not None:
                 scores_list.append(scores_row[key])
                 weights_list.append(weights.get(key, 1.0))
 
@@ -210,36 +209,6 @@ def calc_table_scores(
     return metrics_data
 
 
-def calc_ranks(metrics_data: list[dict]) -> list[dict]:
-    """
-    Calculate rank for each model and add to table data.
-
-    Parameters
-    ----------
-    metrics_data
-        Rows data containing model name, metric values, and Score.
-        The "Score" column is used to calculate the rank, with the highest score ranked
-        1.
-
-    Returns
-    -------
-    list[dict]
-        Rows of data with rank for each model added.
-    """
-    # If a score is None, set to NaN for ranking purposes, but do not rank
-    ranked_scores = rankdata(
-        [x["Score"] if x.get("Score") is not None else np.nan for x in metrics_data],
-        nan_policy="omit",
-        method="max",
-    )
-    for i, row in enumerate(metrics_data):
-        if np.isnan(ranked_scores[i]):
-            row["Rank"] = None
-        else:
-            row["Rank"] = len(ranked_scores) - int(ranked_scores[i]) + 1
-    return metrics_data
-
-
 def get_table_style(
     data: list[TableRow],
     *,
@@ -339,10 +308,7 @@ def rgba_from_val(val: float, vmin: float, vmax: float, cmap: Colormap) -> str:
 
         # Use thresholds
         if normalized:
-            if col != "Rank":
-                min_value, max_value = 1, 0
-            else:
-                min_value, max_value = 1, len(numeric_values)
+            min_value, max_value = 1, 0
         else:
             min_value = min(numeric_values)
             max_value = max(numeric_values)
@@ -370,13 +336,13 @@ def rgba_from_val(val: float, vmin: float, vmax: float, cmap: Colormap) -> str:
     return style_data_conditional
 
 
-def update_score_rank_style(
+def update_score_style(
     data: list[MetricRow],
     weights: dict[str, float] | None = None,
     thresholds: Thresholds | None = None,
 ) -> tuple[list[MetricRow], list[TableRow]]:
     """
-    Update table scores, ranks, and table styles.
+    Update table scores and table styles.
 
     Parameters
     ----------
@@ -395,7 +361,6 @@ def update_score_rank_style(
     """
     weights = clean_weights(weights)
     data = calc_table_scores(data, weights, thresholds)
-    data = calc_ranks(data)
     scored_data = calc_metric_scores(data, thresholds)
     style = get_table_style(data, scored_data=scored_data)
     return data, style

ml_peg/app/build_app.py

@@ -11,11 +11,11 @@
 from dash.html import H1, H3, Div
 from yaml import safe_load
 
-from ml_peg.analysis.utils.utils import calc_ranks, calc_table_scores, get_table_style
+from ml_peg.analysis.utils.utils import calc_table_scores, get_table_style
 from ml_peg.app import APP_ROOT
 from ml_peg.app.utils.build_components import build_weight_components
 from ml_peg.app.utils.register_callbacks import register_benchmark_to_category_callback
-from ml_peg.app.utils.utils import calculate_column_widths, rank_format, sig_fig_format
+from ml_peg.app.utils.utils import calculate_column_widths, sig_fig_format
 from ml_peg.models.get_models import get_model_names
 from ml_peg.models.models import current_models
 
@@ -194,17 +194,13 @@ def build_summary_table(
         data.append({"MLIP": mlip} | summary_data[mlip])
 
     data = calc_table_scores(data)
-    data = calc_ranks(data)
 
-    columns_headers = ("MLIP",) + tuple(tables.keys()) + ("Score", "Rank")
+    columns_headers = ("MLIP",) + tuple(tables.keys()) + ("Score",)
 
     columns = [{"name": headers, "id": headers} for headers in columns_headers]
     for column in columns:
         column_id = column["id"]
-        if column_id == "Rank":
-            column["type"] = "numeric"
-            column["format"] = rank_format()
-        elif column_id != "MLIP":
+        if column_id != "MLIP":
             column["type"] = "numeric"
             column["format"] = sig_fig_format()
 

ml_peg/app/utils/build_components.py

@@ -31,7 +31,7 @@ def grid_template_from_widths(
     widths
         Mapping of column names to pixel widths.
     column_order
-        Ordered metric column names to render between the MLIP, Score, and Rank columns.
+        Ordered metric column names to render between the MLIP and Score columns.
 
     Returns
     -------
@@ -41,7 +41,6 @@ def grid_template_from_widths(
     tracks: list[tuple[str, int]] = [("MLIP", widths["MLIP"])]
     tracks.extend((col, widths[col]) for col in column_order)
     tracks.append(("Score", widths["Score"]))
-    tracks.append(("Rank", widths["Rank"]))
 
     template_parts: list[str] = []
     for _, width in tracks:
@@ -146,7 +145,7 @@ def build_weight_components(
             "Threshold metadata must be provided when use_thresholds=True."
         )
     # Identify metric columns (exclude reserved columns)
-    reserved = {"MLIP", "Score", "Rank", "id"}
+    reserved = {"MLIP", "Score", "id"}
     columns = [col["id"] for col in table.columns if col.get("id") not in reserved]
 
     if not columns:
@@ -367,7 +366,7 @@ def build_test_layout(
 
     # Inline normalization thresholds when metadata is supplied
     if thresholds is not None:
-        reserved = {"MLIP", "Score", "Rank", "id"}
+        reserved = {"MLIP", "Score", "id"}
         metric_columns = [
             col["id"] for col in table.columns if col.get("id") not in reserved
         ]
@@ -663,19 +662,19 @@ def build_threshold_inputs(
             )
         )
 
-    for _ in ("Score", "Rank"):
-        cells.append(
-            Div(
-                "",
-                style={
-                    "width": "100%",
-                    "minWidth": "0",
-                    "maxWidth": "100%",
-                    "boxSizing": "border-box",
-                    "border": "1px solid transparent",
-                },
-            )
+    # Score
+    cells.append(
+        Div(
+            "",
+            style={
+                "width": "100%",
+                "minWidth": "0",
+                "maxWidth": "100%",
+                "boxSizing": "border-box",
+                "border": "1px solid transparent",
+            },
         )
+    )
 
     store = Store(
         id=f"{table_id}-thresholds-store",

ml_peg/app/utils/load.py

@@ -15,7 +15,6 @@
     clean_thresholds,
     clean_weights,
     is_numeric_column,
-    rank_format,
     sig_fig_format,
 )
 
@@ -65,10 +64,7 @@ def rebuild_table(filename: str | Path, id: str) -> DataTable:
         width_labels.append(label_source)
         if column_id is None:
             continue
-        if column_id == "Rank":
-            column["type"] = "numeric"
-            column.setdefault("format", rank_format())
-        elif column.get("type") == "numeric" or is_numeric_column(data, column_id):
+        if column.get("type") == "numeric" or is_numeric_column(data, column_id):
             column["type"] = "numeric"
             column.setdefault("format", sig_fig_format())
         if column_name is not None and not isinstance(column_name, str):

ml_peg/app/utils/register_callbacks.py

@@ -12,7 +12,7 @@
     calc_metric_scores,
     calc_table_scores,
     get_table_style,
-    update_score_rank_style,
+    update_score_style,
 )
 from ml_peg.app.utils.utils import (
     Thresholds,
@@ -67,14 +67,14 @@ def update_summary_table(
                     row[tab] = values[row["MLIP"]]
 
         # Update table contents
-        return update_score_rank_style(summary_data, stored_weights)
+        return update_score_style(summary_data, stored_weights)
 
 
 def register_category_table_callbacks(
     table_id: str, use_thresholds: bool = False
 ) -> None:
     """
-    Register callback to update table scores/rankings when stored values change.
+    Register callback to update table scores when stored values change.
 
     Parameters
     ----------
@@ -218,7 +218,7 @@ def update_table_scores(
             if not table_data:
                 raise PreventUpdate
 
-            scored_rows, style = update_score_rank_style(table_data, stored_weights)
+            scored_rows, style = update_score_style(table_data, stored_weights)
             return scored_rows, style, scored_rows
 
     @callback(
@@ -337,7 +337,7 @@ def update_category_from_benchmark(
             if mlip in benchmark_scores:
                 row[benchmark_column] = benchmark_scores[mlip]
 
-        category_rows, style = update_score_rank_style(category_rows, category_weights)
+        category_rows, style = update_score_style(category_rows, category_weights)
         return category_rows, style, category_rows
 
 

ml_peg/app/utils/utils.py

@@ -52,10 +52,9 @@ def calculate_column_widths(
     # Fixed widths for static columns
     widths.setdefault("MLIP", 150)
     widths.setdefault("Score", 100)
-    widths.setdefault("Rank", 100)
 
     for col in columns:
-        if col not in ("MLIP", "Score", "Rank"):
+        if col not in ("MLIP", "Score"):
             # Calculate width based on column title length
             calculated_width = len(col) * char_width + padding
             # Enforce minimum width
@@ -109,18 +108,6 @@ def sig_fig_format() -> TableFormat.Format:
     )
 
 
-def rank_format() -> TableFormat.Format:
-    """
-    Build a formatter that displays integer ranks.
-
-    Returns
-    -------
-    TableFormat.Format
-        Dash table format configured for integer values.
-    """
-    return TableFormat.Format().scheme(TableFormat.Scheme.decimal_integer)
-
-
 def clean_thresholds(
     raw_thresholds: Mapping[str, Mapping[str, object]] | None,
 ) -> Thresholds:
@@ -271,7 +258,7 @@ def format_metric_columns(
         return None
 
     thresholds = thresholds or {}
-    reserved = {"MLIP", "Score", "Rank", "id"}
+    reserved = {"MLIP", "Score", "id"}
     updated_columns: list[dict[str, object]] = []
 
     for column in columns:
@@ -352,7 +339,7 @@ def format_tooltip_headers(
         return None
 
     thresholds = thresholds or {}
-    reserved = {"MLIP", "Score", "Rank", "id"}
+    reserved = {"MLIP", "Score", "id"}
 
     updated: dict[str, str] = {}
     for key, text in tooltip_header.items():