Update README - GeoFractalDavid-Basin-k50 - Run 20251016_011725 - Acc 68.82%

README.md

@@ -23,7 +23,7 @@ model-index:
       type: imagenet-1k
     metrics:
     - type: accuracy
-      value: 67.78
+      value: 68.82
       name: Validation Accuracy
 ---
 
@@ -34,17 +34,17 @@ Features must "fit" geometric signatures: k-simplex shapes, Cantor positions, an
 
 ## 🎯 Performance
 
-- **Best Validation Accuracy**: 67.78%
-- **Epoch**: 2/10
-- **Training Time**: 4m
+- **Best Validation Accuracy**: 68.82%
+- **Epoch**: 4/10
+- **Training Time**: 8m
 
 ### Per-Scale Performance
-- **Scale 448D**: 65.68%
-- **Scale 512D**: 65.72%
-- **Scale 576D**: 66.88%
-- **Scale 640D**: 65.49%
-- **Scale 704D**: 66.07%
-- **Scale 768D**: 65.25%
+- **Scale 448D**: 62.82%
+- **Scale 512D**: 62.51%
+- **Scale 576D**: 67.01%
+- **Scale 640D**: 61.87%
+- **Scale 704D**: 62.76%
+- **Scale 768D**: 59.67%
 
 
 ## 🏗️ Architecture
@@ -72,8 +72,8 @@ Each scale learns to weight these components differently.
 The alpha parameter controls middle-interval weighting in the Cantor staircase.
 
 - **Initial**: 0.3301
-- **Final**: 0.3377
-- **Change**: +0.0076
+- **Final**: 0.1289
+- **Change**: -0.2012
 - **Converged to 0.5**: False
 
 The Cantor staircase uses soft triadic decomposition with learnable alpha to map
@@ -85,34 +85,34 @@ Each class has a learned scalar Cantor prototype. The model pulls features towar
 their class's Cantor position.
 
 **Scale 448D**:
-- Mean: 0.3299
-- Std: 0.1153
-- Range: [0.0698, 0.5232]
+- Mean: 0.1039
+- Std: 0.0948
+- Range: [-0.0791, 0.3438]
 
 **Scale 512D**:
-- Mean: 0.3303
-- Std: 0.1152
-- Range: [0.0707, 0.5232]
+- Mean: 0.1038
+- Std: 0.0947
+- Range: [-0.0793, 0.3433]
 
 **Scale 576D**:
-- Mean: 0.3406
-- Std: 0.1138
-- Range: [0.0846, 0.5392]
+- Mean: 0.1130
+- Std: 0.0984
+- Range: [-0.0716, 0.3572]
 
 **Scale 640D**:
-- Mean: 0.3284
-- Std: 0.1156
-- Range: [0.0675, 0.5210]
+- Mean: 0.1031
+- Std: 0.0945
+- Range: [-0.0798, 0.3426]
 
 **Scale 704D**:
-- Mean: 0.3376
-- Std: 0.1141
-- Range: [0.0799, 0.5346]
+- Mean: 0.1104
+- Std: 0.0974
+- Range: [-0.0738, 0.3539]
 
 **Scale 768D**:
-- Mean: 0.3321
-- Std: 0.1149
-- Range: [0.0728, 0.5256]
+- Mean: 0.1049
+- Std: 0.0952
+- Range: [-0.0779, 0.3452]
 
 
 Most classes cluster around 0.5 (middle Cantor region), with smooth spread across [0,1].
@@ -122,12 +122,12 @@ This creates a continuous manifold rather than discrete bins.
 
 Each scale learns optimal weights for combining geometric components:
 
-**Scale 448D**: Feature=0.653, Cantor=0.071, Crystal=0.276
-**Scale 512D**: Feature=0.610, Cantor=0.072, Crystal=0.318
-**Scale 576D**: Feature=0.879, Cantor=0.026, Crystal=0.096
-**Scale 640D**: Feature=0.578, Cantor=0.071, Crystal=0.351
-**Scale 704D**: Feature=0.822, Cantor=0.030, Crystal=0.148
-**Scale 768D**: Feature=0.668, Cantor=0.048, Crystal=0.285
+**Scale 448D**: Feature=0.675, Cantor=0.043, Crystal=0.281
+**Scale 512D**: Feature=0.601, Cantor=0.045, Crystal=0.354
+**Scale 576D**: Feature=0.965, Cantor=0.005, Crystal=0.030
+**Scale 640D**: Feature=0.551, Cantor=0.045, Crystal=0.404
+**Scale 704D**: Feature=0.896, Cantor=0.007, Crystal=0.097
+**Scale 768D**: Feature=0.629, Cantor=0.026, Crystal=0.344
 
 
 **Pattern**: Lower scales rely on feature similarity, higher scales use crystal geometry.