Update README - GeoFractalDavid-Basin-k9 - Run 20251016_000149 - Acc 78.89%

README.md

@@ -23,7 +23,7 @@ model-index:
       type: imagenet-1k
     metrics:
     - type: accuracy
-      value: 82.22
+      value: 78.89
       name: Validation Accuracy
 ---
 
@@ -34,20 +34,20 @@ Features must "fit" geometric signatures: k-simplex shapes, Cantor positions, an
 
 ## 🎯 Performance
 
-- **Best Validation Accuracy**: 82.22%
-- **Epoch**: 8/10
-- **Training Time**: 21m
+- **Best Validation Accuracy**: 78.89%
+- **Epoch**: 2/10
+- **Training Time**: 5m
 
 ### Per-Scale Performance
-- **Scale 512D**: 70.80%
-- **Scale 576D**: 70.18%
-- **Scale 640D**: 68.11%
-- **Scale 704D**: 66.96%
-- **Scale 768D**: 80.10%
-- **Scale 832D**: 75.20%
-- **Scale 896D**: 60.24%
-- **Scale 960D**: 59.06%
-- **Scale 1024D**: 53.63%
+- **Scale 512D**: 78.11%
+- **Scale 576D**: 78.22%
+- **Scale 640D**: 78.29%
+- **Scale 704D**: 78.22%
+- **Scale 768D**: 78.10%
+- **Scale 832D**: 78.11%
+- **Scale 896D**: 77.50%
+- **Scale 960D**: 77.48%
+- **Scale 1024D**: 76.10%
 
 
 ## 🏗️ Architecture
@@ -74,10 +74,10 @@ Each scale learns to weight these components differently.
 
 The alpha parameter controls middle-interval weighting in the Cantor staircase.
 
-- **Initial**: 0.4823
-- **Final**: 0.0027
-- **Change**: -0.4796
-- **Converged to 0.5**: False
+- **Initial**: 0.5078
+- **Final**: 0.5387
+- **Change**: +0.0309
+- **Converged to 0.5**: True
 
 The Cantor staircase uses soft triadic decomposition with learnable alpha to map
 features into [0,1] space with fractal structure.
@@ -88,49 +88,49 @@ Each class has a learned scalar Cantor prototype. The model pulls features towar
 their class's Cantor position.
 
 **Scale 512D**:
-- Mean: 0.0732
-- Std: 0.0858
-- Range: [-0.0501, 0.4448]
+- Mean: 0.5389
+- Std: 0.1279
+- Range: [0.2747, 0.7762]
 
 **Scale 576D**:
-- Mean: 0.0729
-- Std: 0.0859
-- Range: [-0.0502, 0.4447]
+- Mean: 0.5391
+- Std: 0.1278
+- Range: [0.2753, 0.7758]
 
 **Scale 640D**:
-- Mean: 0.0732
-- Std: 0.0859
-- Range: [-0.0502, 0.4447]
+- Mean: 0.5391
+- Std: 0.1277
+- Range: [0.2756, 0.7758]
 
 **Scale 704D**:
-- Mean: 0.0729
-- Std: 0.0857
-- Range: [-0.0502, 0.4445]
+- Mean: 0.5384
+- Std: 0.1274
+- Range: [0.2757, 0.7740]
 
 **Scale 768D**:
-- Mean: 0.0756
-- Std: 0.0882
-- Range: [-0.0501, 0.4484]
+- Mean: 0.5344
+- Std: 0.1296
+- Range: [0.2642, 0.7727]
 
 **Scale 832D**:
-- Mean: 0.0751
-- Std: 0.0876
-- Range: [-0.0501, 0.4477]
+- Mean: 0.5376
+- Std: 0.1279
+- Range: [0.2729, 0.7738]
 
 **Scale 896D**:
-- Mean: 0.0735
-- Std: 0.0862
-- Range: [-0.0503, 0.4453]
+- Mean: 0.5361
+- Std: 0.1295
+- Range: [0.2662, 0.7758]
 
 **Scale 960D**:
-- Mean: 0.0731
-- Std: 0.0860
-- Range: [-0.0502, 0.4450]
+- Mean: 0.5367
+- Std: 0.1287
+- Range: [0.2695, 0.7749]
 
 **Scale 1024D**:
-- Mean: 0.0731
-- Std: 0.0859
-- Range: [-0.0502, 0.4450]
+- Mean: 0.5375
+- Std: 0.1283
+- Range: [0.2718, 0.7747]
 
 
 Most classes cluster around 0.5 (middle Cantor region), with smooth spread across [0,1].
@@ -140,15 +140,15 @@ This creates a continuous manifold rather than discrete bins.
 
 Each scale learns optimal weights for combining geometric components:
 
-**Scale 512D**: Feature=0.783, Cantor=0.028, Crystal=0.188
-**Scale 576D**: Feature=0.677, Cantor=0.030, Crystal=0.293
-**Scale 640D**: Feature=0.575, Cantor=0.031, Crystal=0.394
-**Scale 704D**: Feature=0.543, Cantor=0.030, Crystal=0.427
-**Scale 768D**: Feature=0.975, Cantor=0.001, Crystal=0.024
-**Scale 832D**: Feature=0.866, Cantor=0.002, Crystal=0.133
-**Scale 896D**: Feature=0.451, Cantor=0.009, Crystal=0.540
-**Scale 960D**: Feature=0.484, Cantor=0.013, Crystal=0.503
-**Scale 1024D**: Feature=0.262, Cantor=0.009, Crystal=0.729
+**Scale 512D**: Feature=0.692, Cantor=0.072, Crystal=0.235
+**Scale 576D**: Feature=0.651, Cantor=0.073, Crystal=0.276
+**Scale 640D**: Feature=0.619, Cantor=0.074, Crystal=0.307
+**Scale 704D**: Feature=0.626, Cantor=0.066, Crystal=0.308
+**Scale 768D**: Feature=0.829, Cantor=0.031, Crystal=0.140
+**Scale 832D**: Feature=0.694, Cantor=0.048, Crystal=0.258
+**Scale 896D**: Feature=0.802, Cantor=0.032, Crystal=0.166
+**Scale 960D**: Feature=0.735, Cantor=0.038, Crystal=0.227
+**Scale 1024D**: Feature=0.663, Cantor=0.042, Crystal=0.295
 
 
 **Pattern**: Lower scales rely on feature similarity, higher scales use crystal geometry.