From ba2666632981f30f727fbfe28a471431aa7ea628 Mon Sep 17 00:00:00 2001
From: Dominik Roth <dominik.roth.dev@gmail.com>
Date: Sun, 24 Oct 2021 18:07:11 +0200
Subject: [PATCH] Made progress, but not working

---
 train.py | 12 ++++++------
 utils.py | 11 ++++++++---
 2 files changed, 14 insertions(+), 9 deletions(-)

diff --git a/train.py b/train.py
index fc6c9a1..92ccb0a 100644
--- a/train.py
+++ b/train.py
@@ -59,16 +59,16 @@ def evaluate(dataloader, module):
         y_true_batch = y_true_batch.to(device, dtype)  # (batch_size,)
 
         y_pred_batch, p, halting_step = module(x_batch)
-        y_halted_batch = y_pred_batch.gather(
+        y_halted_batch = y_pred_batch.index_select(
             dim=0,
-            index=halting_step[None, :] - 1,
+            index=halting_step - 1,
         )[
             0
-        ]  # (batch_size,)
+        ] # (batch_size,)
 
         # Computing single metrics (mean over samples in the batch)
         accuracy_halted = (
-            ((y_halted_batch > 0) == y_true_batch).to(torch.float32).mean()
+            abs(y_halted_batch - y_true_batch).to(torch.float32).mean()
         )
 
         metrics_single_["accuracy_halted"].append(accuracy_halted)
@@ -78,9 +78,9 @@ def evaluate(dataloader, module):
 
         # Computing per step metrics (mean over samples in the batch)
         accuracy = (
-            ((y_pred_batch > 0) == y_true_batch[None, :])
+            (abs(y_pred_batch - y_true_batch[None, :]))
             .to(torch.float32)
-            .mean(dim=1)
+            .mean(dim=1).mean(dim=1)
         )
 
         metrics_per_step_["accuracy"].append(accuracy)
diff --git a/utils.py b/utils.py
index 27f11da..002f35a 100644
--- a/utils.py
+++ b/utils.py
@@ -102,17 +102,21 @@ class PetriDishNet(nn.Module):
         self.allow_halting = allow_halting
 
         self.synapses = SparseLinear(n_neurons, n_neurons,
-                                     sparsity=0.9,
+                                     sparsity=0.8,
+                                     #sparsity=0.1,
                                      small_world=True,
                                      dynamic=True,
-                                     deltaT=6000,
+                                     #deltaT=6000,
+                                     deltaT=250,
                                      Tend=150000,
+                                     #Tend=5000,
                                      alpha=0.1,
                                      max_size=1e8)
 
         self.ionDucts = ActivationSparsity(alpha=0.1,
                                            beta=1.5,
-                                           act_sparsity=0.65)
+                                           act_sparsity=0.75)
+                                           #act_sparsity=0.10)
 
     def forward(self, x):
         """Run forward pass.
@@ -164,6 +168,7 @@ class PetriDishNet(nn.Module):
             else:
                 lambda_n = torch.sigmoid(state[:, 0])
 
+
             # Store releavant outputs
             y_list.append(state[:, (self.n_in+1):(self.n_in+self.n_out+1)])
             p_list.append(un_halted_prob * lambda_n)