MétaCan
Menu
Back to cohort
Record W7126423241 · doi:10.21428/594757db.269457de

GCE: Confidence Calibration Error for ImprovedTrustworthiness of Graph Neural Networks

2024· article· en· W7126423241 on OpenAlexaff
Hirad Daneshvar, Reza Samavi

Bibliographic record

Venuenot available
Typearticle
Languageen
FieldComputer Science
TopicAdvanced Graph Neural Networks
Canadian institutionsVector InstituteToronto Metropolitan University
Fundersnot available
KeywordsGraphArtificial neural networkConfidence intervalCalibrationHomogeneousPattern recognition (psychology)Code (set theory)Training setError function

Abstract

fetched live from OpenAlex

The popularity of Graph Neural Networks (GNNs) in recent years underscores the need for further exploration of GNNs' trustworthiness. The confidence reported by GNN models in their predictions is an important aspect of trustworthiness, particularly in safety-critical domains such as healthcare. Recent proposals have identified that, unlike other deep learning models, GNNs exhibit under-confidence in their predictions. In this research, we propose Graph Confidence Error (GCE), a loss function to calibrate GNN model confidence during training. We compute the loss by quantifying the contribution of each data point to the model's confidence error and then use this value as a weight parameter in the loss function. We experimentally evaluated our approach for (1) three node classification tasks, including one heterogeneous and two homogeneous graphs, and (2) two graph classification tasks. The evaluation results demonstrate the reduction of the model's anticipated calibration error while preserving its overall performance. The code to GCE is publicly available at this URL https://github.com/samavi/pubs/tree/main/GCE.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.

How this classification was reachedexpand

Full frame machine prediction

Teacher imitation

Not calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.

metaresearch head score (Codex)0.006
metaresearch head score (Gemma)0.053
Version: metacan-v3-hybrid-931329e0061cValidation status: machine_predicted_unvalidated
Candidate categoriesnone
Consensus categoriesnone
DomainCandidate signal: none · Consensus signal: none
Study designCandidate signal: Simulation or modeling · Consensus signal: Simulation or modeling
GenreCandidate signal: Empirical · Consensus signal: none
Teacher disagreement score0.006
Threshold uncertainty score0.032

Distilled classifier scores by category (both heads)

CategoryCodexGemma
Metaresearch0.0060.053
Meta-epidemiology (narrow)0.0020.001
Meta-epidemiology (broad)0.0010.001
Bibliometrics0.0020.001
Science and technology studies0.0010.002
Scholarly communication0.0020.004
Open science0.0030.003
Research integrity0.0020.004
Insufficient payload (model declined to judge)0.0030.001

Machine scores (provisional)

The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.

Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.

Opus teacher head0.016
GPT teacher head0.277
Teacher spread0.260 · how far apart the two teachers sit on this one work
Validation statusscore_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it

Classification

machine, unvalidated

Machine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.

The models applied no category: nothing in the taxonomy fit this work.
Study designSimulation or modeling
Domainnot available
GenreEmpirical

How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".

Quick stats

Citations0
Published2024
Admission routes1
Has abstractyes

Explore more

Same topicAdvanced Graph Neural NetworksFrench-language works237,207