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..
Spectral Convolutional Conditional Neural Processes
Authors: Peiman Mohseni, Nick Duffield
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We validate the effectiveness of SConv CNPs on both synthetic and real-world datasets, demonstrating how ideas from operator learning can advance the capabilities of NPs. 4 Experiments We evaluate our framework on four regression benchmarks and compare its performance against several members of CNPs family. |
| Researcher Affiliation | Academia | Peiman Mohseni Texas A&M University EMAIL Nick Duffield Texas A&M University EMAIL |
| Pseudocode | No | The paper describes methods and processes but does not contain a dedicated section, figure, or block explicitly labeled as 'Pseudocode' or 'Algorithm', nor does it present structured, code-like procedural steps. |
| Open Source Code | Yes | Our implementation and experimental code are publicly available at https://github.com/peiman-m/SConv CNP. |
| Open Datasets | Yes | Traffic Flow For our third experiment, we evaluate on the California traffic-flow dataset from Large ST [Liu et al., 2023b]. This dataset comprises five years (2017 2021) of traffic measurements recorded every 5 minutes by approximately 8,600 loop-detector sensors deployed across California s highway network... available at https://www.kaggle.com/datasets/liuxu77/largest. For our final experiment, we evaluate model performance on an image-completion task formulated as a spatial regression problem... We use images from the Describable Textures Dataset (DTD; Cimpoi et al. [2014]). |
| Dataset Splits | Yes | For validation, we use a fixed set of 4,096 tasks, organized into 256 batches of 16 tasks. In these tasks, the number of query points is fixed at 256, while the number of context points is sampled using the same procedure as during training. Testing follows the same configuration as validation, except that we evaluate on 16,000 test tasks. Training: 5,119 sensors 133,094 windows Validation: 853 sensors 22,178 windows Test: 2,561 sensors 66,586 windows. We adopt the standard DTD split, which contains 1,880 images each for training, validation, and testing. |
| Hardware Specification | Yes | We used a single NVIDIA A100 GPU with 40 GB of memory for all the computations. |
| Software Dependencies | No | All implementations are written in Py Torch [Paszke et al., 2019] and publicly available at https://github.com/peiman-m/SConv CNP. While PyTorch is mentioned, a specific version number for PyTorch itself or other key software components is not provided. |
| Experiment Setup | Yes | Each model is trained for 500 epochs using Adam W [Loshchilov and Hutter, 2017] with a learning rate of 5 10 4. We apply gradient clipping with a maximum norm of 0.5 [Pascanu et al., 2013]. |