Add EP to BP gradient sparsity analysis

EP CIFAR d=256: s(1e-6)=100%, mean_norm=1.41e-04
EP produces networks where ALL samples have non-zero BP gradients,
unlike DFA (0.4%), SB (21%), CB (3%). EP is closer to BP (98.7%).

Updated clean_sparsity_summary.csv (980 rows, now includes EP).

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

1 files changed, 91 insertions, 0 deletions

diff --git a/results/confirmatory/clean_sparsity/cifar_ep_s1024.json b/results/confirmatory/clean_sparsity/cifar_ep_s1024.json
new file mode 100644
index 0000000..65256d9
--- /dev/null
+++ b/results/confirmatory/clean_sparsity/cifar_ep_s1024.json
@@ -0,0 +1,91 @@
+{
+  "method": "ep",
+  "seed": 1024,
+  "d": 256,
+  "per_layer": [
+    {
+      "layer": 0,
+      "mean_norm": 9.15559649001807e-05,
+      "median_norm": 6.607273826375604e-05,
+      "support": {
+        "1e-08": 1.0,
+        "1e-07": 1.0,
+        "1e-06": 1.0,
+        "1e-05": 1.0,
+        "0.0001": 0.3046875
+      },
+      "r_inf_mean": 0.31026577949523926,
+      "pr_mean": 0.1851312220096588,
+      "hoyer_mean": 0.5329517126083374,
+      "topk_energy": {
+        "1": 0.158521369099617,
+        "5": 0.3967384696006775,
+        "10": 0.557002067565918,
+        "25": 0.804891049861908
+      }
+    },
+    {
+      "layer": 1,
+      "mean_norm": 9.101406612899154e-05,
+      "median_norm": 6.56897755106911e-05,
+      "support": {
+        "1e-08": 1.0,
+        "1e-07": 1.0,
+        "1e-06": 1.0,
+        "1e-05": 1.0,
+        "0.0001": 0.30078125
+      },
+      "r_inf_mean": 0.31200894713401794,
+      "pr_mean": 0.18345797061920166,
+      "hoyer_mean": 0.5321497917175293,
+      "topk_energy": {
+        "1": 0.15984469652175903,
+        "5": 0.39870667457580566,
+        "10": 0.5582325458526611,
+        "25": 0.8051235675811768
+      }
+    },
+    {
+      "layer": 2,
+      "mean_norm": 8.733402501093224e-05,
+      "median_norm": 6.450584623962641e-05,
+      "support": {
+        "1e-08": 1.0,
+        "1e-07": 1.0,
+        "1e-06": 1.0,
+        "1e-05": 1.0,
+        "0.0001": 0.2890625
+      },
+      "r_inf_mean": 0.32220906019210815,
+      "pr_mean": 0.17444628477096558,
+      "hoyer_mean": 0.5220228433609009,
+      "topk_energy": {
+        "1": 0.17035308480262756,
+        "5": 0.41159749031066895,
+        "10": 0.5678898692131042,
+        "25": 0.8107988238334656
+      }
+    },
+    {
+      "layer": 3,
+      "mean_norm": 8.715597505215555e-05,
+      "median_norm": 6.27272529527545e-05,
+      "support": {
+        "1e-08": 1.0,
+        "1e-07": 1.0,
+        "1e-06": 1.0,
+        "1e-05": 1.0,
+        "0.0001": 0.28125
+      },
+      "r_inf_mean": 0.32343047857284546,
+      "pr_mean": 0.17165207862854004,
+      "hoyer_mean": 0.5113759636878967,
+      "topk_energy": {
+        "1": 0.17299136519432068,
+        "5": 0.41694819927215576,
+        "10": 0.5739493370056152,
+        "25": 0.8166541457176208
+      }
+    }
+  ]
+}
\ No newline at end of file