# Almost Right: Making First-Layer Kernels Nearly Orthogonal Improves Model Generalization

arXiv: https://arxiv.org/abs/2504.16362 · HTML: https://mlsanity.com/paper/2504.16362/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 28 of its 66 references that are in the map.
- First posted on arXiv 23 Apr 2025.

## Builds on (most cited first, 28 of 28)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020)
- [Swin Transformer: Hierarchical Vision Transformer using Shifted Windows](https://mlsanity.com/paper/2103.14030/) (2021)
- [Densely Connected Convolutional Networks](https://mlsanity.com/paper/1608.06993/) (2016)
- [Analyzing and Improving the Image Quality of StyleGAN](https://mlsanity.com/paper/1912.04958/) (2019)
- [A ConvNet for the 2020s](https://mlsanity.com/paper/2201.03545/) (2022)
- [Progressive Growing of GANs for Improved Quality, Stability, and Variation](https://mlsanity.com/paper/1710.10196/) (2017)
- [Training Generative Adversarial Networks with Limited Data](https://mlsanity.com/paper/2006.06676/) (2020)
- [Alias-Free Generative Adversarial Networks](https://mlsanity.com/paper/2106.12423/) (2021)
- [StarGAN v2: Diverse Image Synthesis for Multiple Domains](https://mlsanity.com/paper/1912.01865/) (2019)
- [Exact solutions to the nonlinear dynamics of learning in deep linear neural networks](https://mlsanity.com/paper/1312.6120/) (2013)
- [All you need is a good init](https://mlsanity.com/paper/1511.06422/) (2015)
- [Orthogonal Convolutional Neural Networks](https://mlsanity.com/paper/1911.12207/) (2019)
- [All You Need is Beyond a Good Init: Exploring Better Solution for Training Extremely Deep Convolutional Neural Networks with Orthonormality and Modulation](https://mlsanity.com/paper/1703.01827/) (2017)
- [Creation and Validation of a Chest X-Ray Dataset with Eye-tracking and Report Dictation for AI Development](https://mlsanity.com/paper/2009.07386/) (2020)
- [A Large-Scale Benchmark for Food Image Segmentation](https://mlsanity.com/paper/2105.05409/) (2021)
- [Optimality of the Johnson-Lindenstrauss Lemma](https://mlsanity.com/paper/1609.02094/) (2016)
- [CYBORG: Blending Human Saliency Into the Loss Improves Deep Learning](https://mlsanity.com/paper/2112.00686/) (2021)
- [LOT: Layer-wise Orthogonal Training on Improving ℓ2 Certified Robustness](https://mlsanity.com/paper/2210.11620/) (2022)
- [Post-mortem Iris Recognition with Deep-Learning-based Image Segmentation](https://mlsanity.com/paper/1901.01708/) (2019)
- [Iris Liveness Detection Competition (LivDet-Iris) -- The 2020 Edition](https://mlsanity.com/paper/2009.00749/) (2020)
- [Human-Aided Saliency Maps Improve Generalization of Deep Learning](https://mlsanity.com/paper/2105.03492/) (2021)
- [Assessment of iris recognition reliability for eyes affected by ocular pathologies](https://mlsanity.com/paper/1809.00206/) (2018)
- [Explain To Me: Salience-Based Explainability for Synthetic Face Detection Models](https://mlsanity.com/paper/2303.11969/) (2023)
- [Teaching AI to Teach: Leveraging Limited Human Salience Data Into Unlimited Saliency-Based Training](https://mlsanity.com/paper/2306.05527/) (2023)
- [MENTOR: Human Perception-Guided Pretraining for Increased Generalization](https://mlsanity.com/paper/2310.19545/) (2023)
- [Grains of Saliency: Optimizing Saliency-based Training of Biometric Attack Detection Models](https://mlsanity.com/paper/2405.00650/) (2024)
- [Explicitly Modeling Pre-Cortical Vision with a Neuro-Inspired Front-End Improves CNN Robustness](https://mlsanity.com/paper/2409.16838/) (2024)

## Abstract (arXiv)

Despite several algorithmic advances in the training of convolutional neural networks (CNNs) over the years, their generalization capabilities are still subpar across several pertinent domains, particularly within open-set tasks often found in biometric and medical contexts. On the contrary, humans have an uncanny ability to generalize to unknown visual stimuli. The efficient coding hypothesis posits that early visual structures (retina, Lateral Geniculate Nucleus, and primary visual cortex) transform inputs to reduce redundancy and maximize information efficiency. This mechanism of redundancy minimization in early vision was the inspiration for CNN regularization techniques that force convolutional kernels to be orthogonal. However, the existing works rely upon matrix projections, architectural modifications, or specific weight initializations, which frequently overtly constrain the network's learning process and excessively increase the computational load during loss function calculation. In this paper, we introduce a flexible and lightweight approach that regularizes a subset of first-layer convolutional filters by making them pairwise-orthogonal, which reduces the redundancy of the extracted features but at the same time prevents putting excessive constraints on the network. We evaluate the proposed method on three open-set visual tasks (anomaly detection in chest X-ray images, synthetic face detection, and iris presentation attack detection) and observe an increase in the generalization capabilities of models trained with the proposed regularizer compared to state-of-the-art kernel orthogonalization approaches. We offer source codes along with the paper.