Search for: All records

Abstract Radiomics provides an appealing, non-invasive approach to probing tumor biology for potential diagnostic and prognostic applications. However, its clinical adoption is limited by challenges in interpretability, which in turn compromise its robustness. To uncover the underlying causation, we developed an ultra-large-scale (ULS) computational model that simulates heterogeneous, vascularized tumor growth under physical constraints to a scale that can be visualized in medical images. Our study revealed the pivotal role of tumor proliferation rate in driving necrosis and tissue heterogeneity and the dominant impact of oxygen consumption rate on vascularization level. Analysis of the resultant tumor Radiomics shows a causal relationship between tumor biophysical parameters and imaging features. Specifically, differences in proliferation and oxygen consumption rates result in distinct changes in radiomic image features, identifying suitable imaging modalities and quantitative imaging metrics for studying these biophysical parameters. We thus reverse-engineer the building blocks of Radiomics as a means to understand their respective biological underpinnings. This work introduces what we believe to be the first computational framework that explicitly links tumor cell biology to macroscopic imaging features-an area traditionally explored through radiomics. 

We consider the problem of fairly allocating the vertices of a graph among n agents, where the value of a bundle is determined by its cut value---the number of edges with exactly one endpoint in the bundle.This model naturally captures applications such as team formation and network partitioning, where valuations are inherently non-monotonic: the marginal values may be positive, negative, or zero depending on the composition of the bundle.We focus on the fairness notion of envy-freeness up to one item (EF1) and explore its compatibility with several efficiency concepts such as Transfer Stability (TS) that prohibits any transfer of item that benefit one agent at another's expense.For general graphs, our results uncover a non-monotonic relationship between the number of agents n and the existence of allocations satisfying EF1 and transfer stability (TS): such allocations always exist for n=2, may fail to exist for n=3, but exist again for all n>= 4. We further show that existence can be guaranteed for any n by slightly weakening the efficiency requirement or by restricting the graph to forests. All of our positive results are achieved via efficient algorithms. 

We introduce Mirage Persistent Kernel (MPK), the first compiler and runtime system that automatically transforms multi-GPU model inference into a single high-performance mega-kernel. MPK introduces an SM-level graph representation that captures data dependencies at the granularity of individual streaming multiprocessors (SMs), enabling cross-operator software pipelining, fine-grained kernel overlap, and other previously infeasible GPU optimizations. The MPK compiler lowers tensor programs into highly optimized SM-level task graphs and generates optimized CUDA implementations for all tasks, while the MPK in-kernel parallel runtime executes these tasks within a single mega-kernel using decentralized scheduling across SMs. Together, these components provide end-to-end kernel fusion with minimal developer effort, while preserving the flexibility of existing programming models. Our evaluation shows that MPK significantly outperforms existing kernel-per-operator LLM serving systems by reducing end-to-end inference latency by up to 1.7×, pushing LLM inference performance close to hardware limits. MPK is publicly available at https://github.com/mirage-project/mirage. 

Liquid crystal elastomers (LCEs) are anisotropic, viscoelastic materials integrating polymer networks and liquid crystals. While their mechanical responses have been extensively studied, their fracture behavior remains largely unexplored. Specifically, the effect of the deformation-director coupling on LCE fracture paths is unknown, and fracture criteria for LCEs are not yet established. To address this gap, we combine experimental and theoretical approaches to investigate fracture propagation in LCEs. We stretch edge-cracked monodomain LCE samples, recording their stress-stretch responses and crack paths under varying initial directors and stretching rates. Our findings reveal that cracks can change direction during propagation, which are highly dependent on both the initial director and the stretching rate. To further understand LCE fracture behavior, we develop a rate-dependent phase-field fracture model, which is validated through experiments, and demonstrates the ability to predict complex fracture paths. Our study paves the way for designing LCEs with enhanced fracture properties, imperative for their future applications. 

Abstract. Hydraulic redistribution (HR), the movement of water via plant root systems that connect soil compartments with different water potential, should influences soil moisture dynamics particularly in water-limited ecosystems. Realistic representation of HR in ecosystem models is essential to improve the ability of these models to predict ecosystem function in dryland regions. In this study, we integrated HR into the Terrestrial ECOsystem model and employed a Bayesian Markov Chain Monte Carlo technique to optimize soil hydraulic parameters and root conductance using four years of soil moisture observations from a piñon-juniper woodland. We found that (i) integrating HR generally improved model prediction of soil moisture during dry periods, particularly in the top 30 cm of the soil profile, where more than 50 % of root biomass exists, mostly during dry periods; (ii) HR increased surface soil moisture by up to 60 % during dry periods; (iii) HR decreased with increasing precipitation magnitude and frequency, however, the length of dry spells between rainfall events also influenced HR rates; and (iv) upward HR in the top 60 cm soil profile became more pronounced as dry conditions progressed, with rates ranging from 0.10 to 0.50 mm d⁻¹. These findings highlight that HR plays a likely role in sustaining soil moisture during extended dry periods and has a limited effect during precipitation events. Future research should investigate the effect of HR on other ecosystem processes, such as net ecosystem exchange of carbon and evapotranspiration under varying climatic conditions. 

Liquid crystal elastomers (LCEs) are emerging actuating materials composed of polymer networks and liquid crystal mesogens. A plateau in the stress-strain curve of LCEs, typical of the semi-soft characteristics, is commonly observed. Although the classical semi-soft model based on compositional fluctuations intends to capture this feature, it does not accurately predict the stress plateau. Moreover, the extended viscoelastic models often lack quantitative comparisons between their theoretical predictions and experimental results. To address these limitations, we phenomenologically modify the semi-soft model, applying it to capture both of the elastic and viscoelastic responses of LCEs. The modified model is further implemented into finite element simulations and used to study intriguing inhomogeneous deformation of LCEs. We demonstrated robust predictions of our model by quantitatively comparing with experimental results.