Knowledge Tree
machine learning models
Also known as: machine learning model
Facts (23)
Sources
A Comprehensive Review of Neuro-symbolic AI for Robustness ... link.springer.com Dec 9, 2025 11 facts
claimResearchers [47] emphasized that evaluating machine learning models with strong adaptive attacks is necessary to uncover vulnerabilities that remain hidden under weaker tests.
claimThe error rate under attack is a robustness metric that measures the proportion of adversarially perturbed inputs that a machine learning model misclassifies.
claimIntervenability is defined as the capacity to actively modify or interact with a machine learning model’s internal mechanisms or representations to influence its outputs in a controlled and predictable manner.
claimTargeted robustness evaluates a machine learning model’s ability to avoid being steered toward a specific, attacker-chosen target label, which complements untargeted notions that only measure any form of misclassification.
claimThe minimum distance, measured using norms like ℓ0, ℓ1, or ℓ2, captures how easily a machine learning model can be manipulated by subtle input changes, where a smaller minimum distance implies higher sensitivity and a greater likelihood of misclassification due to minimal input perturbations.
claimMinimum distance serves as both an attackability measure and a benchmark for evaluating defense mechanisms in machine learning models.
claimPrediction variance measures the stability of a machine learning model's output when subjected to minor perturbations in the input, serving as an indicator of robustness and reliability.
claimMany defense mechanisms in machine learning models fail under properly constructed attacks due to reliance on gradient obfuscation, as demonstrated by [48].
referenceResearch by [51] demonstrated that adversarial examples often lie near decision boundaries, where machine learning models produce low confidence predictions.
claimA confidence score represents a machine learning model's self-assessed certainty in its prediction, typically expressed as the probability assigned to the predicted class.
claimUnder adversarial attacks or distributional shifts, increases in metrics such as predictive entropy and variance indicate a machine learning model's uncertainty and potential struggle to produce reliable predictions.
Knowledge Graphs: Opportunities and Challenges - Springer Nature link.springer.com Apr 3, 2023 2 facts
procedureContent-based recommender systems operate by analyzing the content features of items (such as descriptions or documents) that have been previously scored by target users, and then employing machine learning models to learn user interests.
procedureKnowledge graph completion models train machine learning models on existing graphs to assess the plausibility of new candidate triplets and add those with high plausibility to the graph.
Recent breakthroughs in the valorization of lignocellulosic biomass ... pubs.rsc.org Jun 7, 2025 2 facts
measurementAnwar et al. trained and tested their machine learning models for predicting concrete compressive strength using a dataset of 695 data points, split into 70% training data and 30% testing data, utilizing seven independent variables and five metrics.
measurementIn a research study, 7 out of 10 machine learning models predicted the effectiveness of cellulose nanofiber (CNF) addition in enhancing the compressive strength of concrete materials with an R2-value greater than 0.6.
Understanding LLM Understanding skywritingspress.ca Jun 14, 2024 1 fact
claimMachine learning models are typically assumed to be rational agents that seek the highest probability and lowest cost account of the provided data.
Addressing common challenges with knowledge graphs - SciBite scibite.com 1 fact
claimMachine learning models can be trained using the curated output from the TExpress tool to identify relationships between entities within specific contexts.
A Survey on the Theory and Mechanism of Large Language Models arxiv.org Mar 12, 2026 1 fact
referenceThe paper 'DOGE: domain reweighting with generalization estimation' introduces a method for domain reweighting to improve generalization in machine learning models.
Medicinal plants and human health: a comprehensive review of ... link.springer.com Nov 5, 2025 1 fact
claimMachine learning models can predict environmental conditions or genetic modifications that enhance the production of specific therapeutic compounds in plants, thereby guiding research and reducing the need for trial-and-error experimentation.
Cyber Insights 2025: Open Source and Software Supply Chain ... securityweek.com Jan 15, 2025 1 fact
claimAI and machine learning models rely on dynamic and often opaque supply chains, where each machine learning component, data set, and algorithm may introduce unique vulnerabilities.
Overcoming the limitations of Knowledge Graphs for Decision ... xpertrule.com 1 fact
claimComposite AI can handle complex decision-making tasks more effectively than Knowledge Graphs by combining the strengths of decision trees, machine learning models, and other AI techniques.
Designing Knowledge Graphs for AI Reasoning, Not Guesswork linkedin.com Jan 14, 2026 1 fact
claimPiers Fawkes asserts that current Large Language Models (LLMs) fail to provide sufficient depth because they flatten tables into text, while task-specific machine learning models fail to provide sufficient breadth because they are built one use case at a time.
Track: Poster Session 3 - aistats 2026 virtual.aistats.org 1 fact
claimRegularization methods intended to mitigate shortcuts in machine learning models can sometimes overregularize, inadvertently suppressing causal features along with spurious ones.