Network Bending: Expressive Manipulation of Generative Models in Multiple Domains

This paper presents the network bending framework, a new approach for manipulating and interacting with deep generative models. We present a comprehensive set of deterministic transformations that can be inserted as distinct layers into the computational graph of a trained generative neural network and applied during inference. In addition, we present a novel algorithm for analysing the deep generative model and clustering features based on their spatial activation maps. This allows features to be grouped together based on spatial similarity in an unsupervised fashion. This results in the meaningful manipulation of sets of features that correspond to the generation of a broad array of semantically significant features of the generated results. We outline this framework, demonstrating our results on deep generative models for both image and audio domains. We show how it allows for the direct manipulation of semantically meaningful aspects of the generative process as well as allowing for a broad range of expressive outcomes.

Direct Deviations in Astrocyte Free Ca2+ Concentration Control Multiple Arteriole Tone States

Astrocytes elicit bidirectional control of microvascular diameter in acutely isolated brain slices through vasoconstriction and vasodilation pathways that can be differentially recruited via the free Ca2+ concentration in endfeet and/or the metabolic status of the tissue. However, the Ca2+-level hypothesis has not been tested using direct manipulation. To overcome this, we used Ca2+-clamp whole-cell patching of peri-arteriole astrocytes to change astrocyte-free Ca2+ to different concentrations and examined the vascular response. We discovered that clamping Ca2+ at the approximate resting value (100 nM) had no impact on arteriole diameter in a pre-constricted arteriole. However, a moderate elevation to 250 nM elicited sustained vasodilation that was blocked by the COX-1 antagonist SC-560 (500 nM). The vasodilation to 250 nM Ca2+ was sensitive to the metabolic state, as it converted to vasoconstriction when oxygen tension was dramatically elevated. In normal oxygen, clamping astrocyte Ca2+ well above the resting level (750 nM) produced sustained vasoconstriction, which converted to vasodilation in the 20-HETE blocker HET0016 (1 μM). This response was fully blocked by the addition of SC-560 (500 nM), showing that 20-HETE-induced vasoconstriction dominated the dilatory action of COX-1. These data demonstrate that direct changes in astrocyte free Ca2+ can control multiple arteriole tone states through different mediators.

Retooling Microbiome Engineering for a Sustainable Future

Microbial communities (microbiomes) have been harnessed in biotechnology applications such as wastewater treatment and bioremediation for over a century. Traditionally, engineering approaches have focused on shaping the environment to steer microbiome function versus direct manipulation of the microbiome’s metabolic network.