Search for: All records

Although accuracy and computation benchmarks are widely available to help choose among neural network models, these are usually trained on datasets with many classes, and do not give a good idea of performance for few (<10) classes. The conventional procedure to predict performance involves repeated training and testing on the different models and dataset variations. We propose an efficient cosine similarity-based classification difficulty measure S that is calculated from the number of classes and intra- and inter-class similarity metrics of the dataset. After a single stage of training and testing per model family, relative performance for different datasets and models of the same family can be predicted by comparing difficulty measures – without further training and testing. Our proposed method is verified by extensive experiments on 8 CNN and ViT models and 7 datasets. Results show that S is highly correlated to model accuracy with correlation coefficient r=0.796, outperforming the baseline Euclidean distance at r=0.66. We show how a practitioner can use this measure to help select an efficient model 6 to 29x faster than through repeated training and testing. We also describe using the measure for an industrial application in which options are identified to select a model 42% smaller than the baseline YOLOv5-nano model, and if class merging from 3 to 2 classes meets requirements, 85% smaller. 

Networks and dense suspensions frequently reside near a boundary between soft (or fluidlike) and rigid (or solidlike) regimes. Transitions between these regimes can be driven by changes in structure, density, or applied stress or strain. In general, near the onset or loss of rigidity in these systems, dissipation-limiting heterogeneous nonaffine rearrangements dominate the macroscopic viscoelastic response, giving rise to diverging relaxation times and power-law rheology. Here, we describe a simple quantitative relationship between nonaffinity and the excess viscosity. We test this nonaffinity-viscosity relationship computationally and demonstrate its rheological consequences in simulations of strained filament networks and dense suspensions. We also predict critical signatures in the rheology of semiflexible and stiff biopolymer networks near the strain stiffening transition. 

Flash caches are used to reduce peak backend load for throughput-constrained data center services, reducing the total number of backend servers required. Bulk storage systems are a large-scale example, backed by high-capacity but low-throughput hard disks, and using flash caches to provide a more cost-effective storage layer underlying everything from blobstores to data warehouses. However, flash caches must address the limited write endurance of flash by limiting the long-term average flash write rate to avoid premature wearout. To do so, most flash caches must use admission policies to filter cache insertions and maximize the workload-reduction value of each flash write. The Baleen flash cache uses coordinated ML admission and prefetching to reduce peak backend load. After learning painful lessons with our early ML policy attempts, we exploit a new cache residency model (which we call episodes) to guide model training. We focus on optimizing for an end-to-end system metric (Disk-head Time) that measures backend load more accurately than IO miss rate or byte miss rate. Evaluation using Meta traces from seven storage clusters shows that Baleen reduces Peak Disk-head Time (and hence the number of backend hard disks required) by 12% over state-of-the-art policies for a fixed flash write rate constraint. Baleen-TCO, which chooses an optimal flash write rate, reduces our estimated total cost of ownership (TCO) by 17%. Code and traces are available at https://www.pdl.cmu.edu/CILES/. 

Asynchronously replicated primary-backup databases are commonly deployed to improve availability and offload read-only transactions. To both apply replicated writes from the primary and serve read-only transactions, the backups implement a cloned concurrency control protocol. The protocol ensures read-only transactions always return a snapshot of state that previously existed on the primary. This compels the backup to exactly copy the commit order resulting from the primary's concurrency control. Existing cloned concurrency control protocols guarantee this by limiting the backup's parallelism. As a result, the primary's concurrency control executes some workloads with more parallelism than these protocols. In this paper, we prove that this parallelism gap leads to unbounded replication lag, where writes can take arbitrarily long to replicate to the backup and which has led to catastrophic failures in production systems. We then design C5, the first cloned concurrency protocol to provide bounded replication lag. We implement two versions of C5: Our evaluation in MyRocks, a widely deployed database, demonstrates C5 provides bounded replication lag. Our evaluation in Cicada, a recent in-memory database, demonstrates C5 keeps up with even the fastest of primaries. 

Abstract Reconstructing the behavior of extinct species is challenging, particularly for those with no living analogues. However, damage preserved as paleopathologies on bone can record how an animal moved in life, potentially reflecting behavioral patterns. Here, we assess hypothesized etiologies of pathology in a pelvis and associated right femur of aSmilodon fatalissaber-toothed cat, one of the best-studied species from the Pleistocene-age Rancho La Brea asphalt seeps, California, USA, using visualization by computed tomography (CT). The pelvis exhibits massive destruction of the right hip socket that was interpreted, for nearly a century, to have developed from trauma and infection. CT imaging reveals instead that the pathological distortions characterize chronic remodeling that began at birth and led to degeneration of the joint over the animal’s life. These results suggest that this individual suffered from hip dysplasia, a congenital condition common in domestic dogs and cats. This individual reached adulthood but could not have hunted properly nor defended territory on its own, likely relying on a social group for feeding and protection. While extant social felids are rare, these fossils and others with similar pathologies are consistent with a spectrum of social strategies inSmilodonsupported by a predominance of previous studies. 

Abstract Critical cancer pathways often cannot be targeted because of limited efficiency crossing cell membranes. Here we report the development of a Salmonella-based intracellular delivery system to address this challenge. We engineer genetic circuits that (1) activate the regulatorflhDCto drive invasion and (2) induce lysis to release proteins into tumor cells. Released protein drugs diffuse from Salmonella containing vacuoles into the cellular cytoplasm where they interact with their therapeutic targets. Control of invasion withflhDCincreases delivery over 500 times. The autonomous triggering of lysis after invasion makes the platform self-limiting and prevents drug release in healthy organs. Bacterial delivery of constitutively active caspase-3 blocks the growth of hepatocellular carcinoma and lung metastases, and increases survival in mice. This success in targeted killing of cancer cells provides critical evidence that this approach will be applicable to a wide range of protein drugs for the treatment of solid tumors.