Results 61 to 70 of about 19,051 (188)
InforMask: Unsupervised Informative Masking for Language Model Pretraining [PDF]
arXiv, 2022 Masked language modeling is widely used for pretraining large language models for natural language understanding (NLU). However, random masking is suboptimal, allocating an equal masking rate for all tokens. In this paper, we propose InforMask, a new unsupervised masking strategy for training masked language models.
arxiv
Masks Fusion with Multi-Target Learning For Speech Enhancement [PDF]
arXiv, 2021 Recently, deep neural network (DNN) based time-frequency (T-F) mask estimation has shown remarkable effectiveness for speech enhancement. Typically, a single T-F mask is first estimated based on DNN and then used to mask the spectrogram of noisy speech in an order to suppress the noise.
arxiv
Alternative Protein Sources: Addressing Global Food Security and Environmental Sustainability
Sustainable Development, EarlyView.ABSTRACT The projected global population increase to 8.60 billion by 2050 and 11.20 billion by 2100 underscores the urgent need for sustainable solutions to secure essential food resources, particularly protein, for human health. Traditional animal‐based proteins, while nutritionally rich, are environmentally costly due to greenhouse gas emissions ...
Nilesh Nirmal +4 more
wiley +1 more source
Toxins, 2021 The natural co-occurrence of 42 mycotoxins was investigated in unprocessed oat grains grown in Ireland. The sample set included a total of 208 oat crops harvested during 2015–2016 and produced using conventional, organic, or gluten free farming systems ...
Lorenzo De Colli +8 more
doaj +1 more source
Should You Mask 15% in Masked Language Modeling? [PDF]
arXiv, 2022 Masked language models (MLMs) conventionally mask 15% of tokens due to the belief that more masking would leave insufficient context to learn good representations; this masking rate has been widely used, regardless of model sizes or masking strategies. In this work, we revisit this important choice of MLM pre-training.
arxiv
Feed‐induced hypersalivation in horses from Austria, Germany and Switzerland
Equine Veterinary Journal, EarlyView.Abstract Background While previous reports come mostly from the southern Americas, several outbreaks of hypersalivation in horses were observed in Middle Europe from 2016 to 2018. Objective To describe feed‐induced hypersalivation in European horses. Study design Analysis of feedstuffs.
Linda Franziska Böswald +5 more
wiley +1 more source
Learning Better Masking for Better Language Model Pre-training [PDF]
arXiv, 2022 Masked Language Modeling (MLM) has been widely used as the denoising objective in pre-training language models (PrLMs). Existing PrLMs commonly adopt a Random-Token Masking strategy where a fixed masking ratio is applied and different contents are masked by an equal probability throughout the entire training.
arxiv
Plant Breeding, EarlyView.ABSTRACT Disease resistance traits are complex and quantitative in nature. Breeders regularly evaluate multiple important traits across diverse environments to employ them in genomics‐assisted breeding. In this study, we evaluated the prospects of genomic prediction models by incorporating genome‐wide association study (GWAS) results into single‐trait ...
Vinay Kumar Reddy Nannuru +9 more
wiley +1 more source
Frontiers in Microbiology, 2019 Cereals can be contaminated by several mycotoxins, whose co-presence may represent an undervalued risk for humans and animals. Maize and wheat are the most contaminated cereals and in temperate areas could be affected in field conditions by several ...
Valentina Scarpino +2 more
doaj +1 more source
Mask to reconstruct: Cooperative Semantics Completion for Video-text Retrieval [PDF]
arXiv, 2023 Recently, masked video modeling has been widely explored and significantly improved the model's understanding ability of visual regions at a local level. However, existing methods usually adopt random masking and follow the same reconstruction paradigm to complete the masked regions, which do not leverage the correlations between cross-modal content ...
arxiv