Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
Rethinking Hebbian Principle: Low-Dimensional Structural Projection for Unsupervised Learning
Authors: Shikuang Deng, Jiayuan Zhang, Yuhang Wu, Ting Chen, Shi Gu
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experimental results show that SPHe Re achieves SOTA performance among unsupervised synaptic plasticity approaches on standard image classification benchmarks, including CIFAR-10, CIFAR-100, and Tiny-Image Net. Furthermore, the method exhibits strong effectiveness in continual learning and transfer learning scenarios, and image reconstruction tasks show the robustness and generalizability of the extracted features. |
| Researcher Affiliation | Academia | 1School of Computer Science and Engineering, UESTC 2Glasgow College, UESTC 3 School of Computer Science and Technology, Zhejiang University 4 State Key Lab of Brain-Machine Intelligence, Zhejiang University, China EMAIL, EMAIL |
| Pseudocode | No | The paper describes the methodology in prose and mathematical equations in sections 4 and 4.4, accompanied by a conceptual diagram in Figure 1, but does not include any explicitly labeled 'Pseudocode' or 'Algorithm' blocks. |
| Open Source Code | Yes | Our code is available at https://github.com/brain-intelligence-lab/SPHe Re. |
| Open Datasets | Yes | Extensive experimental results show that SPHe Re achieves SOTA performance among unsupervised synaptic plasticity approaches on standard image classification benchmarks, including CIFAR-10, CIFAR-100, and Tiny-Image Net. |
| Dataset Splits | Yes | For data splits and hyperparameters, we choose the universal standard setting. |
| Hardware Specification | Yes | All the experiments are carried out on one Nvidia RTX 3090 Graphic Card with 24GB VRAM. The detailed information is in below: Table 14: Computational resources used for experiments. Training was conducted on an NVIDIA RTX 3090 GPU (24GB VRAM) with a batch size of 128 and dataloader workers set to 2. |
| Software Dependencies | No | The paper mentions using Adam W optimizer and Leaky-Re LU activation function, but does not provide specific version numbers for any key software components or libraries like Python, PyTorch, or TensorFlow. |
| Experiment Setup | Yes | We trained the network using Adam W (learning rate: 0.001, weight decay: 0.05) with a batch size of 128 and a learning rate scheduler; the hyperparameter λ for Lorth is set to 0.8; no data augmentation was applied except for standard normalization. |