Federated Learning (FL) enables learning a shared model across many clients without violating the privacy requirements. One of the key attributes in FL is the heterogeneity that exists in both resource and data due to the differences in computation and communication capacity, as well as the quantity and content of data among different clients. We conduct a case study to show that heterogeneity in resource and data has a significant impact on training time and model accuracy in conventional FL systems. To this end, we propose TiFL, a Tier-based Federated Learning System, which divides clients into tiers based on their training performance and selects clients from the same tier in each training round to mitigate the straggler problem caused by heterogeneity in resource and data quantity. To further tame the heterogeneity caused by non-IID (Independent and Identical Distribution) data and resources, TiFL employs an adaptive tier selection approach to update the tiering on-the-fly based on the observed training performance and accuracy overtime. We prototype TiFL in a FL testbed following Google's FL architecture and evaluate it using popular benchmarks and the state-of-the-art FL benchmark LEAF. Experimental evaluation shows that TiFL outperforms the conventional FL in various heterogeneous conditions. With the proposed adaptive tier selection policy, we demonstrate that TiFL achieves much faster training performance while keeping the same (and in some cases - better) test accuracy across the board.
Metastases are responsible for the majority of cancer-related deaths. Although genomic heterogeneity within primary tumors is associated with relapse, heterogeneity among treatment-naïve metastases has not been comprehensively assessed. We analyzed sequencing data for 76 untreated metastases from 20 patients and inferred cancer phylogenies for breast, colorectal, endometrial, gastric, lung, melanoma, pancreatic, and prostate cancers. We found that within individual patients, a large majority of driver gene mutations are common to all metastases. Further analysis revealed that the driver gene mutations that were not shared by all metastases are unlikely to have functional consequences.
Piezoelectric materials are important as sensors and transducers for applications such as ultrasonics. Liu et al. discovered nanopillar regions in a sodium-niobium-oxide film that substantially improve the piezoelectric properties (see the Perspective by Bassiri-Gharb). These nanopillar regions reverse where the cations and anions are located in the crystal structure, with a distinctive boundary in between. This difference in structure results in a strain-sensitive polarity that enhances the piezoelectric properties in a chemically simple material. Science this issue p. ; see also p.  High-performance piezoelectric materials are critical components for electromechanical sensors and actuators. For more than 60 years, the main strategy for obtaining large piezoelectric response has been to construct multiphase boundaries, where nanoscale domains with local structural and polar heterogeneity are formed, by tuning complex chemical compositions. We used a different strategy to emulate such local heterogeneity by forming nanopillar regions in perovskite oxide thin films. We obtained a giant effective piezoelectric coefficient d33,f* of ~1098 picometers per volt with a high Curie temperature of ~450°C. Our lead-free composition of sodium-deficient sodium niobate contains only three elements (Na, Nb, and O). The formation of local heterogeneity with nanopillars in the perovskite structure could be the basis for a general approach to designing and optimizing various functional materials. : /lookup/doi/10.1126/science.abb3209 : /lookup/doi/10.1126/science.abc8007
We propose a novel class of network models for temporal dyadic interaction data. Our objective is to capture important features often observed in social interactions: sparsity, degree heterogeneity, community structure and reciprocity. We use mutually-exciting Hawkes processes to model the interactions between each (directed) pair of individuals. The intensity of each process allows interactions to arise as responses to opposite interactions (reciprocity), or due to shared interests between individuals (community structure). For sparsity and degree heterogeneity, we build the non time dependent part of the intensity function on compound random measures following Todeschini et al., 2016.
Since complex diseases such as cancer, diabetes and so on pose a very big threat to human health, they have been extensively studied in the past decades1. However, the underlying pathogenesis of complex diseases is still not clearly known. With the rapid development of genomics technologies, the big data of variations on DNA level such as SNP and CNV (copy number variation) allow comprehensive characterization of complex diseases and provide potential biomarkers to predict the status of complex diseases. Due to the'missing heritability' and lack of reproducibility, the exploration of relationships between SNPs and complex diseases have been transferred from single variation to biomarkers interactions which are defined as epistasis2. First, as the number of variants increases, the combination space expands exponentially, resulting in the'curse of dimensionality' problem.