B-Cos Alignment for Inherently Interpretable CNNs and Vision Transformers.

We present a new direction for increasing the interpretability of deep neural networks (DNNs) by promoting weight-input alignment during training. For this, we propose to replace the linear transformations in DNNs by our novel B-cos transformation. As we show, a sequence (network) of such transformations induces a single linear transformation that faithfully summarises the full model computations.

Better Understanding Differences in Attribution Methods via Systematic Evaluations.

Deep neural networks are very successful on many vision tasks, but hard to interpret due to their black box nature. To overcome this, various post-hoc attribution methods have been proposed to identify image regions most influential to the models' decisions. Evaluating such methods is challenging since no ground truth attributions exist.

Impact of selenium biofortification on production characteristics of forages grown following standard management practices in Oregon.

Introduction: Low selenium (Se) concentrations in soils and plants pose a health risk for ruminants consuming locally-grown forages. Previous studies have shown that Se concentrations in forages can be increased using soil-applied selenate amendments. However, the effects of foliar selenate amendments applied with traditional nitrogen-phosphorus-potassium-sulfur (NPKS) fertilizers on forage yields, and nutrient contents, and agronomic efficiencies are unknown.

Effects of Amount and Chemical Form of Selenium Amendments on Forage Selenium Concentrations and Species Profiles.

Selenium (Se) agronomic biofortification of plants is effective for alleviating Se deficiencies in human and livestock populations. Less is known about how higher selenate amendment rates, or how foliar compared with granular selenate amendments affect forage Se concentrations. Therefore, we compared the effects of a higher sodium selenate foliar amendment rate (900 vs.

Optimising for Interpretability: Convolutional Dynamic Alignment Networks.

We introduce a new family of neural network models called Convolutional Dynamic Alignment Networks (CoDA Nets), which are performant classifiers with a high degree of inherent interpretability. Their core building blocks are Dynamic Alignment Units (DAUs), which are optimised to transform their inputs with dynamically computed weight vectors that align with task-relevant patterns. As a result, CoDA Nets model the classification prediction through a series of input-dependent linear transformations, allowing for linear decomposition of the output into individual input contributions.