method
active
method:neuron-resamplingNeuron Resampling
Periodically reinitializing dead autoencoder neurons using high-loss data points to improve feature coverage
Neighborhood — ranked by edge-count
Concepts (1)
concept
- Dead Neuronsassociated_withAutoencoder neurons that fail to activate across any datapoints during training; addressed via neuron resampling
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- Michael Johnson's prior work on how neural networks (and brains) can be 'annealed' to find optimal states.
- Signaling molecules that move under bioelectric control; shared between neural and developmental contexts.
- A neuron that responds to multiple unrelated inputs, posing a major challenge for circuit-level interpretation
- Method used to identify and partition the 28 MLP neurons into disjoint clusters by Fourier period
- Brain-based physical implementations of consciousness-related functions, assumed by many ToCs to be exclusive.
- Key insight that rotating a neural representation to a non-standard basis can reveal distributed causal structure invisible in standard neuron-aligned basis.