MOTIVATION: The Variant Call Format (VCF) is widely used in genome sequencing but scales poorly. For instance, we estimate a 150 000 genome VCF would occupy 900 TiB, making it costly and complicated to produce, analyze, and store. The issue stems from VCF's requirement to densely represent both reference-genotypes and allele-indexed arrays. These requirements lead to unnecessary data duplication and, ultimately, very large files. RESULTS: To address these challenges, we introduce the Scalable Variant Call Representation (SVCR). This representation reduces file sizes by ensuring they scale linearly with samples. SVCR's linear scaling relies on two techniques, both necessary for linearity: local allele indices and reference blocks, which were first introduced by the Genomic Variant Call Format. SVCR is also lossless and mergeable, allowing for N + 1 and N + K incremental joint-calling. We present two implementations of SVCR: SVCR-VCF, which encodes SVCR in VCF format, and VDS, which uses Hail's native format. Our experiments confirm the linear scalability of SVCR-VCF and VDS, in contrast to the super-linear growth seen with standard VCF files. We also discuss the VDS Combiner, a scalable, open-source tool for producing a VDS from GVCFs and unique features of VDS which enable rapid data analysis. SVCR, and VDS in particular, ensure the scientific community can generate, analyze, and disseminate genetics datasets with millions of samples. AVAILABILITY AND IMPLEMENTATION: https://github.com/hail-is/hail/.

We present P5; an event-driven policy framework that allows network operators to realize end-to-end policies on top of P4-based data planes in an intuitive and effective manner. We demonstrate how P5 adheres to a service-level agreement (SLA) by applying P4-based traffic engineering with latency constraints.

Although BERT-style encoder models are heavily used in NLP research, many researchers do not pretrain their own BERTs from scratch due to the high cost of training. In the past half-decade since BERT first rose to prominence, many advances have been made with other transformer architectures and training configurations that have yet to be systematically incorporated into BERT. Here, we introduce MosaicBERT, a BERT-style encoder architecture and training recipe that is empirically optimized for fast pretraining. This efficient architecture incorporates FlashAttention, Attention with Linear Biases (ALiBi), Gated Linear Units (GLU), a module to dynamically remove padded tokens, and low precision LayerNorm into the classic transformer encoder block. The training recipe includes a 30% masking ratio for the Masked Language Modeling (MLM) objective, bfloat16 precision, and vocabulary size optimized for GPU throughput, in addition to best-practices from RoBERTa and other encoder models. When pretrained from scratch on the C4 dataset, this base model achieves a downstream average GLUE (dev) score of 79.6 in 1.13 hours on 8 A100 80 GB GPUs at a cost of roughly $20. We plot extensive accuracy vs. pretraining speed Pareto curves and show that MosaicBERT base and large are consistently Pareto optimal when compared to a competitive BERT base and large. This empirical speed up in pretraining enables researchers and engineers to pretrain custom BERT-style models at low cost instead of finetune on existing generic models. We open source our model weights and code.

Casual games are a serious business, reaching a wide audience of players and generating billions of dollars in revenue, but serious games have struggled to find a market breakthrough. Considerable scholarly work has been published on principles for both serious game design and casual game design, but less research has been done on how to integrate these two approaches in order to realize the advantageous qualities of casual games in serious game projects. “Thinking casual” holds potential for making serious game projects both more feasible and more impactful. This paper introduces the term “activist-casual” to characterize a design framework based on the goals of conscientious design and confluences between serious and casual game design principles. The proposed activist-casual framework comprises ten hybrid principles, as well as five conceptual guardrails against the harmful tendencies of the commercial casual game industry. This approach could help activist designers overcome the challenge of limited development resources and create meaningfully impactful games.

Deterministic and time-sensitive networking will be critical for enabling ultra-reliable low latency communication (URLLC) networks and services. However, limitations exist for using traditional IP technologies and current Internet architecture, for deterministic and time-sensitive networking. The emerging URLLC network types and traffic types require new fundamental cornerstones that include deterministic behaviour, real-time sensitivity, and capable of being highly resilient to network infrastructure failures. Various technologies exist with the current Internet to provide limited capabilities for each primitive. However, fully achieving the URLLC objectives will require a radical approach that moves beyond current network design and Internet architecture.This paper proposes a new approach and architecture for extending the data plane and control plane elements for future URLCC requirements. It provides a summary of applicable transport, control and service layers. How existing functions and mechanisms will need to be extended and combined to meet the future demands of Future Internet networks and services. Finally, the paper provides a theoretical PCE-based framework for future deterministic and time-sensitive networks, and the protocol work required to standardize and deploy.

Identification of new concepts in scientific literature can help power faceted search, scientific trend analysis, knowledge-base construction, and more, but current methods are lacking. Manual identification can't keep up with the torrent of new publications, while the precision of existing automatic techniques is too low for many applications. We present an unsupervised concept extraction method for scientific literature that achieves much higher precision than previous work. Our approach relies on a simple but novel intuition: each scientific concept is likely to be introduced or popularized by a single paper that is disproportionately cited by subsequent papers mentioning the concept. From a corpus of computer science papers on arXiv, we find that our method achieves a [email protected] of 99%, compared to 86% for prior work, and a substantially better precision-yield trade-off across the top 15,000 extractions. To stimulate research in this area, we release our code and data.

Arman Cohan, Iz Beltagy, Daniel King, Bhavana Dalvi, Dan Weld. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP). 2019.

Network operators recognise that Content Delivery Networks are essential for meeting user Internet application and content demands. The infrastructure must be tightly integrated to provide request routing, content caching, load balancing, scalability and reliability, whilst minimising deployment time and complexity. A major step towards achieving these goals is to embrace recent Software Defined Network and Network Function Virtualisation objectives and design principles. This paper outlines the OpenCache API; an interface used to define the behaviour and operation of an SDN-based content delivery platform in real-time. We demonstrate the applicability and effectiveness of such an API by implementing load-balancing and fail-over functionalities as part of an experimental deployment.