Publications

Fluorescence lifetime is a relatively new contrast mechanism for optical imaging in living subjects that relies on intrinsic properties of fluorophores rather than concentration dependent intensity. Drawing upon the success of fluorescence lifetime imaging microscopy (FLIM) for investigation of protein-protein interactions and intracellular physiology, in vivo fluorescence lifetime imaging (FLI) promises to dramatically increase the utility of fluorescencebased imaging in preclinical and clinical applications. Intrinsic fluorescence lifetime measurements in living tissues can distinguish pathologies such as cancer from healthy tissue. Unfortunately, intrinsic FLT contrast is limited to superficial measurements. Conventional intensity-based agents have been reported for measuring these phenomena in vitro, but translation into living animals is difficult due to optical properties of tissues. For this reason, contrast agents that can be detected in the near infrared (NIR) wavelengths are being developed by our lab and others to enhance the capabilities of this modality. FLT is less affected by concentration and thus is better for detecting small changes in physiology, as long as sufficient fluorescence signal can be measured. FLT can also improve localization of signals for improved deep tissue imaging. Examples of the utility of exogenous contrast agents will be discussed, including applications in monitoring physiologic functions, controlled drug release and cancer biology. Instrumentation for FLI will also be discussed, including planar and diffuse optical imaging in time and frequency domains. Future applications will also be discussed that are being developed in this exciting field that complement other optical modalities.

Neural programs are highly accurate and structured policies that perform algorithmic tasks by controlling the behavior of a computation mechanism. Despite the potential to increase the interpretability and the compositionality of the behavior of artificial agents, it remains difficult to learn from demonstrations neural networks that represent computer programs. The main challenges that set algorithmic domains apart from other imitation learning domains are the need for high accuracy, the involvement of specific structures of data, and the extremely limited observability. To address these challenges, we propose to model programs as Parametrized Hierarchical Procedures (PHPs). A PHP is a sequence of conditional operations, that uses a program counter, along with the observation, to select between taking an elementary action, invoking another PHP as a sub-procedure, and returning to the caller. We develop an algorithm for training PHPs from a mixture of annotated and unannotated demonstrations, and apply it to efficient level-wise training of multi-level PHPs. We show in two benchmarks, NanoCraft and long-hand addition, that PHPs can learn neural programs more accurately from smaller amounts of strong and weak supervision.

The Arabidopsis mutant pho1 is deficient in the transfer of Pi from root epidermal and cortical cells to the xylem. The PHO1 gene was identified by a map-based cloning strategy. The N-terminal half of PHO1 is mainly hydrophilic, whereas the C-terminal half has six potential membrane-spanning domains. PHO1 shows no homology with any characterized solute transporter, including the family of H(+)-Pi cotransporters identified in plants and fungi. PHO1 shows highest homology with the Rcm1 mammalian receptor for xenotropic murine leukemia retroviruses and with the Saccharomyces cerevisiae Syg1 protein involved in the mating pheromone signal transduction pathway. PHO1 is expressed predominantly in the roots and is upregulated weakly under Pi stress. Studies with PHO1 promoter-beta-glucuronidase constructs reveal predominant expression of the PHO1 promoter in the stelar cells of the root and the lower part of the hypocotyl. There also is beta-glucuronidase staining of endodermal cells that are adjacent to the protoxylem vessels. The Arabidopsis genome contains 10 additional genes showing homology with PHO1. Thus, PHO1 defines a novel class of proteins involved in ion transport in plants.

The mechanical properties of the bcc refractory high-entropy alloy (RHEA) NbMoTaW with columnar and equiaxed microstructures and a nanoscale metal oxide layer on the grain boundaries are investigated at ambient temperatures (RT) to 1200 °C. Under compression, the alloy shows a yield strength, σ y, of ⁓1390 ± 20 MPa at RT and retains a high yield strength, σ y, of ⁓301.5 MPa at 1200 °C. However, in tension, the fracture strengths, σ f, of both columnar and equiaxed microstructures are far lower - below 70 MPa - and the material fails in the elastic regime without showing any measurable ductility at all temperatures. The fracture mechanisms in tension transition from transgranular cleavage at RT to a mixture of transgranular and intergranular fracture at 800 °C to a complete intergranular fracture at 1200 °C due to selective weakening of the metal oxide layer on the grain boundaries driven by a structural change. The transition in the fracture mechanism in the equiaxed microstructure promotes extrinsic toughening at higher temperatures, resulting in a ⁓6.4 times higher fracture toughness (K Ic ⁓10.3 MPa√m) at 1200 °C compared to that at RT. NbMoTaW is a well-known RHEA, but the poor tensile strength and fracture toughness measured in this study highlight the critical importance of investigating the mechanical properties of these high-temperature RHEAs in tension.

As Internet video goes mainstream, we see increasing user expectations for higher video quality and new global policy requirements for content providers. Inspired by the case for centralizing network-layer control, we present C3, a control system for optimizing Internet video delivery. The design of C3 addresses key challenges in ensuring scalability and tackling data plane heterogeneity. First, to ensure scalability and responsiveness, C3 introduces a novel split control plane architecture that can tolerate a small increase in model staleness for a dramatic increase in scalability. Second, C3 supports diverse client-side platforms via a minimal clientside sensing/actuation layer and offloads complex monitoring and control logic to the control plane. C3 has been operational for eight years, and today handles more than 100M sessions per day from 244 countries for 100+ content providers and has improved the video quality significantly. In doing so, C3 serves as a proof point of the viability of fine-grained centralized control for Internetscale applications. Our experiences reinforce the case for centralizing control with the continued emergence of new use case pulls (e.g., client diversity) and technology pushes (e.g., big data platforms).

Field-sensitive electrochemical reactions are controlled by electrode charging which is sensitive to the size of the electrolyte containing cations.

Abstract Electrochemical noise (EN) sensors have been developed to measure the corrosion rate of Type 304 steel (SS) in subcritical and supercritical environments. The potential and coupling current noise were recorded simultaneously and the noise resistance (Rn) was calculated from the standard deviations in the potential and current records. We found that the inverse noise resistance correlated very well with the corrosion rate evaluated from separate mass loss experiments, and that both the inverse noise resistance and the average corrosion rate were functions of temperature and flow rate. In the temperature range from 200 to 390°C, the corrosion rate was found to be proportional to the inverse noise resistance and hence the Stern-Geary relationship can be used to evaluate the corrosion rate.