Fixes, bug updates and pip upgrades

.gitignore

@@ -23,4 +23,5 @@ dist/*
 logs/*
 .pytest_cache/*
 .vscode/*
+.claude/
 data/*

bindsnet/evaluation/evaluation.py

@@ -25,7 +25,8 @@ def assign_labels(
         ``assign_labels()`` call.
     :param alpha: Rate of decay of label assignments.
     :return: Tuple of class assignments, per-class spike proportions, and per-class
-        firing rates.
+        firing rates. Neurons that never fired are assigned ``-1`` (unassigned) so
+        they do not bias predictions toward class ``0``.
     """
 
     n_neurons = spikes.size(2)
@@ -35,10 +36,10 @@ def assign_labels(
 
     # Sum over time dimension (spike ordering doesn't matter).
     spikes = spikes.sum(1)
-    
+
     for i in range(n_labels):
         # Create mask.
-        mask = (labels == i)
+        mask = labels == i
         # Count the number of samples with this label.
         n_labeled = mask.sum().float()
 
@@ -50,24 +51,23 @@ def assign_labels(
 
     # Compute proportions of spike activity per class.
     total_activity = rates.sum(1, keepdim=True)
-    proportions = torch.where(total_activity > 0, rates / total_activity, torch.zeros_like(rates))
+    proportions = torch.where(
+        total_activity > 0, rates / total_activity, torch.zeros_like(rates)
+    )
 
     # Noise for random tie breaking.
-    eps = 1e-6 # Small enough not to distort real decisions
+    eps = 1e-6  # Small enough not to distort real decisions
     noise = eps * torch.randn_like(proportions)
-    
+
     # Neuron assignments are the labels they fire most for.
     assignments = torch.argmax(proportions + noise, dim=1)
 
-    # Uniform assignment for silent neurons
-    silent_mask = total_activity.squeeze() == 0
-    n_silent = silent_mask.sum()
+    # Neurons that never fired have no class preference; mark them with -1 so
+    # they are masked out of downstream voting (issue #736: silent neurons must
+    # not default to a vote for class 0). Downstream schemes only match
+    # ``assignments == i`` for i >= 0, so -1 neurons are excluded.
+    assignments[rates.sum(1) == 0] = -1
 
-    if n_silent > 0:
-        assignments[silent_mask] = torch.randint(
-            0, n_labels, (n_silent,), device=spikes.device
-        )
-
     return assignments, proportions, rates
 
 
@@ -118,7 +118,9 @@ def all_activity(
     :param assignments: A vector of shape ``(n_neurons,)`` of neuron label assignments.
     :param n_labels: The number of target labels in the data.
     :return: Predictions tensor of shape ``(n_samples,)`` resulting from the "all
-        activity" classification scheme.
+        activity" classification scheme. Samples that elicit no activity from any
+        assigned neuron are predicted as ``-1`` (abstain) rather than defaulting to
+        class ``0``.
     """
     n_samples = spikes.size(0)
 
@@ -140,7 +142,12 @@ def all_activity(
             rates[:, i] = torch.sum(spikes[:, indices], 1) / n_assigns
 
     # Predictions are arg-max of layer-wise firing rates.
-    return torch.sort(rates, dim=1, descending=True)[1][:, 0]
+    predictions = torch.sort(rates, dim=1, descending=True)[1][:, 0]
+
+    # Abstain (-1) on samples with no activity, avoiding a biased vote for class 0.
+    predictions[rates.sum(1) == 0] = -1
+
+    return predictions
 
 
 def proportion_weighting(
@@ -161,7 +168,9 @@ def proportion_weighting(
         proportions of neuron spiking activity.
     :param n_labels: The number of target labels in the data.
     :return: Predictions tensor of shape ``(n_samples,)`` resulting from the "proportion
-        weighting" classification scheme.
+        weighting" classification scheme. Samples that elicit no weighted activity from
+        any assigned neuron are predicted as ``-1`` (abstain) rather than defaulting to
+        class ``0``.
     """
     n_samples = spikes.size(0)
 
@@ -187,6 +196,9 @@ def proportion_weighting(
     # Predictions are arg-max of layer-wise firing rates.
     predictions = torch.sort(rates, dim=1, descending=True)[1][:, 0]
 
+    # Abstain (-1) on samples with no activity, avoiding a biased vote for class 0.
+    predictions[rates.sum(1) == 0] = -1
+
     return predictions