Update README - GeoFractalDavid-Basin-k9 - Run 20251015_233559 - Acc 78.94%

README.md

@@ -14,7 +14,7 @@ metrics:
 - accuracy
 library_name: pytorch
 model-index:
-- name: GeoFractalDavid-Basin-k4
+- name: GeoFractalDavid-Basin-k9
   results:
   - task:
       type: image-classification
@@ -23,29 +23,31 @@ model-index:
       type: imagenet-1k
     metrics:
     - type: accuracy
-      value: 82.36
+      value: 78.94
       name: Validation Accuracy
 ---
 
-# GeoFractalDavid-Basin-k4: Geometric Basin Classification
+# GeoFractalDavid-Basin-k9: Geometric Basin Classification
 
 **GeoFractalDavid** achieves classification through geometric compatibility rather than cross-entropy.
 Features must "fit" geometric signatures: k-simplex shapes, Cantor positions, and hierarchical structure.
 
 ## 🎯 Performance
 
-- **Best Validation Accuracy**: 82.36%
-- **Epoch**: 10/10
-- **Training Time**: 26m 54s
+- **Best Validation Accuracy**: 78.94%
+- **Epoch**: 2/10
+- **Training Time**: 5m
 
 ### Per-Scale Performance
-- **Scale 576D**: 69.88%
-- **Scale 640D**: 69.78%
-- **Scale 704D**: 67.68%
-- **Scale 768D**: 81.33%
-- **Scale 832D**: 65.97%
-- **Scale 896D**: 70.34%
-- **Scale 960D**: 59.12%
+- **Scale 512D**: 78.05%
+- **Scale 576D**: 78.18%
+- **Scale 640D**: 78.17%
+- **Scale 704D**: 78.07%
+- **Scale 768D**: 77.93%
+- **Scale 832D**: 77.51%
+- **Scale 896D**: 77.29%
+- **Scale 960D**: 76.83%
+- **Scale 1024D**: 76.89%
 
 
 ## 🏗️ Architecture
@@ -55,9 +57,9 @@ Features must "fit" geometric signatures: k-simplex shapes, Cantor positions, an
 **Core Components**:
 - **Feature Dimension**: 768
 - **Number of Classes**: 1000
-- **k-Simplex Structure**: k=4 (5 vertices per class)
-- **Scales**: [576, 640, 704, 768, 832, 896, 960]
-- **Total Simplex Vertices**: 5,000
+- **k-Simplex Structure**: k=9 (10 vertices per class)
+- **Scales**: [512, 576, 640, 704, 768, 832, 896, 960, 1024]
+- **Total Simplex Vertices**: 10,000
 
 **Geometric Components**:
 1. **Feature Similarity**: Cosine similarity to k-simplex centroids
@@ -72,9 +74,9 @@ Each scale learns to weight these components differently.
 
 The alpha parameter controls middle-interval weighting in the Cantor staircase.
 
-- **Initial**: 0.4817
-- **Final**: 0.0247
-- **Change**: -0.4570
+- **Initial**: 0.4823
+- **Final**: 0.4067
+- **Change**: -0.0756
 - **Converged to 0.5**: False
 
 The Cantor staircase uses soft triadic decomposition with learnable alpha to map
@@ -85,40 +87,50 @@ features into [0,1] space with fractal structure.
 Each class has a learned scalar Cantor prototype. The model pulls features toward
 their class's Cantor position.
 
+**Scale 512D**:
+- Mean: 0.4160
+- Std: 0.1368
+- Range: [0.0494, 0.7292]
+
 **Scale 576D**:
-- Mean: 0.0753
-- Std: 0.1164
-- Range: [-0.0998, 0.4649]
+- Mean: 0.4171
+- Std: 0.1361
+- Range: [0.0492, 0.7291]
 
 **Scale 640D**:
-- Mean: 0.0751
-- Std: 0.1163
-- Range: [-0.0999, 0.4648]
+- Mean: 0.4178
+- Std: 0.1356
+- Range: [0.0497, 0.7288]
 
 **Scale 704D**:
-- Mean: 0.0753
-- Std: 0.1164
-- Range: [-0.0999, 0.4648]
+- Mean: 0.4183
+- Std: 0.1352
+- Range: [0.0495, 0.7274]
 
 **Scale 768D**:
-- Mean: 0.0787
-- Std: 0.1191
-- Range: [-0.0950, 0.4677]
+- Mean: 0.4165
+- Std: 0.1294
+- Range: [0.0494, 0.6725]
 
 **Scale 832D**:
-- Mean: 0.0750
-- Std: 0.1163
-- Range: [-0.0999, 0.4647]
+- Mean: 0.4172
+- Std: 0.1301
+- Range: [0.0493, 0.6754]
 
 **Scale 896D**:
-- Mean: 0.0771
-- Std: 0.1179
-- Range: [-0.0978, 0.4663]
+- Mean: 0.4170
+- Std: 0.1329
+- Range: [0.0493, 0.6931]
 
 **Scale 960D**:
-- Mean: 0.0761
-- Std: 0.1170
-- Range: [-0.0991, 0.4654]
+- Mean: 0.4173
+- Std: 0.1335
+- Range: [0.0495, 0.6973]
+
+**Scale 1024D**:
+- Mean: 0.4171
+- Std: 0.1341
+- Range: [0.0497, 0.7015]
 
 
 Most classes cluster around 0.5 (middle Cantor region), with smooth spread across [0,1].
@@ -128,13 +140,15 @@ This creates a continuous manifold rather than discrete bins.
 
 Each scale learns optimal weights for combining geometric components:
 
-**Scale 576D**: Feature=0.791, Cantor=0.028, Crystal=0.182
-**Scale 640D**: Feature=0.741, Cantor=0.029, Crystal=0.230
-**Scale 704D**: Feature=0.640, Cantor=0.031, Crystal=0.329
-**Scale 768D**: Feature=0.991, Cantor=0.001, Crystal=0.008
-**Scale 832D**: Feature=0.613, Cantor=0.026, Crystal=0.362
-**Scale 896D**: Feature=0.772, Cantor=0.002, Crystal=0.226
-**Scale 960D**: Feature=0.252, Cantor=0.004, Crystal=0.743
+**Scale 512D**: Feature=0.688, Cantor=0.075, Crystal=0.237
+**Scale 576D**: Feature=0.646, Cantor=0.075, Crystal=0.279
+**Scale 640D**: Feature=0.609, Cantor=0.074, Crystal=0.317
+**Scale 704D**: Feature=0.599, Cantor=0.069, Crystal=0.332
+**Scale 768D**: Feature=0.855, Cantor=0.027, Crystal=0.118
+**Scale 832D**: Feature=0.800, Cantor=0.032, Crystal=0.168
+**Scale 896D**: Feature=0.704, Cantor=0.041, Crystal=0.255
+**Scale 960D**: Feature=0.667, Cantor=0.045, Crystal=0.289
+**Scale 1024D**: Feature=0.633, Cantor=0.046, Crystal=0.321
 
 
 **Pattern**: Lower scales rely on feature similarity, higher scales use crystal geometry.