Integrated Calendar

The emergence of “ultra-strength materials” that can withstand significant fractions of the ideal strength at component scale without any inelastic relaxation harbingers a new field within mechanics of materials. Recently, we have experimentally achieved more than 13% reversible tensile strains in Si that fundamentally redefines what it means to be Si, and ~7% uniform tensile strain in micron-scale single-crystalline diamond bridge arrays, where thousands of transistors and quantum sensors can be integrated in one diamond microbridge. Elastic Strain Engineering (ESE) aims to endow material structures with very large stresses and stress gradients to guide electronic, photonic, and spin excitations and control energy, mass, and information flows. As “smaller is stronger” for most engineering materials at room temperature, a much larger dynamical range of tensile-and-shear deviatoric stresses for small-scale structures can be achieved, as the defect (e.g., dislocation, crack) population dynamics change from defect-propagation to defect-nucleation controlled. Thus, all six stress components can be used to tune the physical and chemical properties of a material like a 7-element alloy. Four pillars of ESE need to be addressed experimentally and computationally: (a) making materials and structures that can withstand deviatoric elastic strain patterns that are exceptionally large and extended in space-time volume, inhomogeneous, dynamically reversible, or combinations thereof, (b) measuring and understanding how functional properties such as photonic and electronic characteristics vary with imposed elastic strain tensor, (c) characterizing and modeling the mechanisms of stress relaxations; the goal is not to use them for forming but to defeat them at service temperatures (usually room temperature and above) and extended timescales, and (d) computational design based on first principles, e.g. predicting ideal strength surface, topological changes in band structures, etc. assisted by machine learning.

Working memory (WM), the process through which information is transiently maintained and manipulated over a brief period of time, is essential for most cognitive functions. However, the mechanisms underlying the generation and stability of WM neuronal representations at the population level remain elusive. To uncover these mechanisms, we trained head-fixed mice to perform  an olfactory working memory task and used optogenetics to delineate circuits causal for behavioral performance. We used mesoscopic and light bead  two photon imaging to record from up to 35,000 secondary motor cortical neurons simulataneously across multiple days and show differential stabilization of different task parameters with learning and practice of the task. We find that cortical working memory representations causal for task performance are highly volatile but only stabilize after multiple days of practice well after task learning. We hypothesize that representational drift soon after learning may allow for higher levels of flexibility for new task rules. 
I will also review some of the new open-source tools developed for large-scale imaging of neural activity patterns in freely behaving animals.

Text-conditioned image editing has recently attracted considerable interest. However, most methods are currently either limited to specific editing types (e.g., object overlay, style transfer), or apply to synthetically generated images, or require multiple input images of a common object. In this paper we demonstrate, for the very first time, the ability to apply complex (e.g., non-rigid) text-guided semantic edits to a single real image. For example, we can change the posture and composition of one or multiple objects inside an image, while preserving its original characteristics. Our method can make a standing dog sit down or jump, cause a bird to spread its wings, etc. — each within its single high-resolution natural image provided by the user. Contrary to previous work, our proposed method requires only a single input image and a target text (the desired edit). It operates on real images, and does not require any additional inputs (such as image masks or additional views of the object). Our method, which we call "Imagic", leverages a pre-trained text-to-image diffusion model for this task. It produces a text embedding that aligns with both the input image and the target text, while fine-tuning the diffusion model to capture the image-specific appearance. We demonstrate the quality and versatility of our method on numerous inputs from various domains, showcasing a plethora of high quality complex semantic image edits, all within a single unified framework