# Integrating domain knowledge: using hierarchies to improve deep classifiers

arXiv: https://arxiv.org/abs/1811.07125 · HTML: https://mlsanity.com/paper/1811.07125/

## Key facts

- Cited by 25 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 17 Nov 2018.

## Built on by (most cited first, 25 of 25)

- [ImageNet-21K Pretraining for the Masses](https://mlsanity.com/paper/2104.10972/) (2021)
- [Learning Hierarchy Aware Features for Reducing Mistake Severity](https://mlsanity.com/paper/2207.12646/) (2022)
- [No Cost Likelihood Manipulation at Test Time for Making Better Mistakes in Deep Networks](https://mlsanity.com/paper/2104.00795/) (2021)
- [Integrating Language Guidance into Vision-based Deep Metric Learning](https://mlsanity.com/paper/2203.08543/) (2022)
- [Visually Consistent Hierarchical Image Classification](https://mlsanity.com/paper/2406.11608/) (2024)
- [Hierarchical classification at multiple operating points](https://mlsanity.com/paper/2210.10929/) (2022)
- [Inducing Neural Collapse to a Fixed Hierarchy-Aware Frame for Reducing Mistake Severity](https://mlsanity.com/paper/2303.05689/) (2023)
- [Revisiting Hierarchical Text Classification: Inference and Metrics](https://mlsanity.com/paper/2410.01305/) (2024)
- [From GNNs to Trees: Multi-Granular Interpretability for Graph Neural Networks](https://mlsanity.com/paper/2505.00364/) (2025)
- [United We Learn Better: Harvesting Learning Improvements From Class Hierarchies Across Tasks](https://mlsanity.com/paper/2107.13627/) (2021)
- [WikiChurches: A Fine-Grained Dataset of Architectural Styles with Real-World Challenges](https://mlsanity.com/paper/2108.06959/) (2021)
- [All Mistakes Are Not Equal: Comprehensive Hierarchy Aware Multi-label Predictions (CHAMP)](https://mlsanity.com/paper/2206.08653/) (2022)
- [Integrating Visual and Semantic Similarity Using Hierarchies for Image Retrieval](https://mlsanity.com/paper/2308.08431/) (2023)
- [Hypernym Bias: Unraveling Deep Classifier Training Dynamics through the Lens of Class Hierarchy](https://mlsanity.com/paper/2502.12125/) (2025)
- [Transition Matrix Representation of Trees with Transposed Convolutions](https://mlsanity.com/paper/2202.10677/) (2022)
- [Loss Functions for Classification using Structured Entropy](https://mlsanity.com/paper/2206.07122/) (2022)
- [TaxoKnow: Taxonomy as Prior Knowledge in the Loss Function of Multi-class Classification](https://mlsanity.com/paper/2305.16341/) (2023)
- [Priority-Aware Clinical Pathology Hierarchy Training for Multiple Instance Learning](https://mlsanity.com/paper/2507.20469/) (2025)
- [A neurosymbolic Approach with Epistemic Deep Learning for Hierarchical Image Classification](https://mlsanity.com/paper/2605.16383/) (2026)
- [Self-Supervised Learning from Semantically Imprecise Data](https://mlsanity.com/paper/2104.10901/) (2021)
- [Computing Class Hierarchies from Classifiers](https://mlsanity.com/paper/2112.01187/) (2021)
- [Introducing the ICBe Dataset: Very High Recall and Precision Event Extraction from Narratives about International Crises](https://mlsanity.com/paper/2202.07081/) (2022)
- [Descriptive Image-Text Matching with Graded Contextual Similarity](https://mlsanity.com/paper/2505.09997/) (2025)
- [Concept Flow Models: Anchoring Concept-Based Reasoning with Hierarchical Bottlenecks](https://mlsanity.com/paper/2606.19489/) (2026)
- [Label Granularity Skew in Federated Learning with Hierarchical Image Classification](https://mlsanity.com/paper/2608.09236/) (2026)

## Abstract (arXiv)

One of the most prominent problems in machine learning in the age of deep learning is the availability of sufficiently large annotated datasets. For specific domains, e.g. animal species, a long-tail distribution means that some classes are observed and annotated insufficiently. Additional labels can be prohibitively expensive, e.g. because domain experts need to be involved. However, there is more information available that is to the best of our knowledge not exploited accordingly. In this paper, we propose to make use of preexisting class hierarchies like WordNet to integrate additional domain knowledge into classification. We encode the properties of such a class hierarchy into a probabilistic model. From there, we derive a novel label encoding and a corresponding loss function. On the ImageNet and NABirds datasets our method offers a relative improvement of 10.4% and 9.6% in accuracy over the baseline respectively. After less than a third of training time, it is already able to match the baseline's fine-grained recognition performance. Both results show that our suggested method is efficient and effective.