Exploring the role of texture features in deep convolutional neural networks: Insights from Portilla-Simoncelli statistics.

It is well-understood that the performance of Deep Convolutional Neural Networks (DCNNs) in image recognition tasks is influenced not only by shape but also by texture information. Despite this, understanding the internal representations of DCNNs remains a challenging task. This study employs a simplified version of the Portilla-Simoncelli Statistics, termed "minPS," to explore how texture information is represented in a pre-trained VGG network. Using minPS features extracted from texture images, we perform a sparse regression on the activations across various channels in VGG layers. Our findings reveal that channels in the early to middle layers of the VGG network can be effectively described by minPS features. Additionally, we observe that the explanatory power of minPS sub-groups evolves as one ascends the network hierarchy. Specifically, sub-groups termed Linear Cross Scale (LCS) and Energy Cross Scale (ECS) exhibit weak explanatory power for VGG channels. To investigate the relationship further, we compare the original texture images with their synthesized counterparts, generated using VGG, in terms of minPS features. Our results indicate that the absence of certain minPS features suggests their non-utilization in VGG's internal representations.

Tuning the geometries of a de novo blue copper protein by axial interactions.

The axial interactions of Cu(2+) in type 1 copper proteins control the physical characteristics of the proteins. We tuned the geometries of a de novo designed blue copper protein with a four-helical bundle structure. The designed protein axially bound various ligands, such as chloride, phosphate, sulfate, acetate, azide, and imidazole, to Cu(2+), exhibiting a blue or green color. The UV-vis spectral bands were observed at approximately 600 nm and approximately 450 nm, with the A (~450)/A (~600) ratios between 0.14 and 1.58. The stronger axial interaction shifted the geometry of the type 1 copper site from trigonal planar geometry (blue copper) toward a tetrahedral-like geometry (green copper). Resonance Raman spectral analyses showed that the phosphate-bound type had the highest-strength Cu-S bond, similar to that of plastocyanin. The chloride-bound type exhibited features similar to those of stellacyanin and nitrite reductase, and the imidazole-bound type exhibited features similar to those of azurin M121E mutant.