The paper introduces an algebraic framework that links neural networks to neural ideals, providing algorithms for computing and approximating these ideals. It demonstrates how to identify and interpret the features captured by each hidden‑layer neuron, validated on the MNIST dataset. An interactive software tool is released to visualize these neuron‑specific features.
arXiv:2609.06862v1 Announce Type: new Abstract: Superposition refers to neural networks representing more features than they have dimensions. It offers a possible explanation for polysemantic neurons...
arXiv:2310.16295v2 Announce Type: replace-cross Abstract: Neural network have achieved remarkable successes in many scientific fields. However, the interpretability of the neural network model is sti...
arXiv:2607. 23397v1 Announce Type: new Abstract: Hierarchical neural networks are widely used in artificial intelligence, yet their mathematical properties remain incompletely understood.
arXiv:2405. 02369v2 Announce Type: replace-cross Abstract: In the past decade, many successful networks are on novel architectures, which almost exclusively use the same type of neurons.
arXiv:2605. 06938v2 Announce Type: replace-cross Abstract: Recently Brown et al.
arXiv:2311. 02960v5 Announce Type: replace Abstract: Over the past decade, deep learning has proven to be a highly effective tool for learning meaningful features from raw data.
The paper introduces a flexible symbolic framework that efficiently computes logical explanations for deep neural networks by parameterizing explanations with internal neuron activations and leveraging general-purpose logical engines like SMT solvers. Unlike previous methods that rely on specialized verifiers or are limited to individual input features, this approach is not restricted in shape and can scale to deep architectures. Experiments on image recognition and medical benchmarks demonstrate improved computational efficiency and the ability to explain networks that were previously intractable for logic-based methods.
arXiv:2501.18018v2 Announce Type: replace-cross Abstract: The neurons of artificial neural networks were originally invented when much less was known about biological neurons than is known today. Our...
The paper investigates how multilayer perceptrons (MLPs) learn features in regression tasks with clustered data. It finds that instead of forming a single global low‑dimensional representation, MLPs develop monosemantic specialized neurons—each neuron aligns strongly with a specific predictive feature relevant to a particular region of the input space. This specialization results in a collection of local low‑dimensional representations, giving MLPs a provable data‑efficiency advantage over methods that rely on a global representation.
arXiv:2607. 07035v1 Announce Type: cross Abstract: The architecture of deep feedforward neural networks is ubiquitous in deep learning, either as a whole system or as a subnetwork of other architectures, and thus its mechanism is a key ingredient of the black box of neural networks.
arXiv:2608.29530v1 Announce Type: cross Abstract: Modern systems in artificial intelligence (AI) somehow excel in domains for which they seem poorly suited. Intelligence has traditionally been modele...
arXiv:2607. 22979v1 Announce Type: cross Abstract: Deep neural networks have been widely used in many applications (e.