An official website of the United States government
With the advancement and dominant service of Internet videos, the content-based video deduplication system becomes an essential and dependent infrastructure for Internet video service. However, the explosively growing video data on the Internet challenges the system design and implementation for its scalability in several ways. (1) Although the quantization-based indexing techniques are effective for searching visual features at a large scale, the costly re-training over the complete dataset must be done periodically. (2) The high-dimensional vectors for visual features demand increasingly large SSD space, degrading I/O performance. (3) Videos crawled from the Internet are diverse, and visually similar videos are not necessarily the duplicates, increasing deduplication complexity. (4) Most videos are edited ones. The duplicate contents are more likely discovered as clips inside the videos, demanding processing techniques with close attention to details. To address above-mentioned issues, we propose Maze, a full-fledged video deduplication system. Maze has an ANNS layer that indexes and searches the high dimensional feature vectors. The architecture of the ANNS layer supports efficient reads and writes and eliminates the data migration caused by re-training. Maze adopts the CNN-based feature and the ORB feature as the visual features, which are optimized for the specific video deduplication task. The features are compact and fully reside in the memory. Acoustic features are also incorporated in Maze so that the visually similar videos but having different audio tracks are recognizable. A clip-based matching algorithm is developed to discover duplicate contents at a fine granularity. Maze has been deployed as a production system for two years. It has indexed 1.3 billion videos and is indexing ~800 thousand videos per day. For the ANNS layer, the average read latency is 4 seconds and the average write latency is at most 4.84 seconds. The re-training over the complete dataset is no longer required no matter how many new data sets are added, eliminating the costly data migration between nodes. Maze recognizes the duplicate live streaming videos with both the similar appearance and the similar audio at a recall of 98%. Most importantly, Maze is also cost-effective. For example, the compact feature design helps save 5800 SSDs and the computation resources devoted to running the whole system decrease to 250K standard cores per billion videos.
more »
« less
A-maze of Natural Stories: Comprehension and surprisal in the Maze task
Behavioral measures of word-by-word reading time provide experimental evidence to test theories of language processing. A-maze is a recent method for measuring incremental sentence processing that can localize slowdowns related to syntactic ambiguities in individual sentences. We adapted A-maze for use on longer passages and tested it on the Natural Stories corpus. Participants were able to comprehend these longer text passages that they read via the Maze task. Moreover, the Maze task yielded useable reaction time data with word predictability effects that were linearly related to surprisal, the same pattern found with other incremental methods. Crucially, Maze reaction times show a tight relationship with properties of the current word, with little spillover of effects from previous words. This superior localization is an advantage of Maze compared with other methods. Overall, we expanded the scope of experimental materials, and thus theoretical questions, that can be studied with the Maze task.
more »
« less
- Award ID(s):
- 2121074
- PAR ID:
- 10488153
- Publisher / Repository:
- eScholarship
- Date Published:
- Journal Name:
- Glossa Psycholinguistics
- Volume:
- 2
- Issue:
- 1
- ISSN:
- 2767-0279
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Theta-burst stimulation (TBS) is a repetitive transcranial magnetic stimulation technique that can be used to upregulate or downregulate different brain regions. However, the timing of its effects and the differing effects of continuous TBS (cTBS) versus intermittent TBS (iTBS) in the reading system have not been explored. This study assessed how stimulation type and post-stimulation timing affected change in performance during a phonological discrimination and sight word recognition task after stimulation of supramarginal gyrus (SMG). Fourteen right-handed young adults (age 18–27 years; 44% male) were block-randomized to receive either iTBS or cTBS to the supramarginal gyrus. Participants then performed a pseudoword discrimination task and an orthographic awareness task (behavioral control) at four different time points and change in reaction time compared to baseline was measured from each time point. There was no effect of stimulation type on change in reaction time [t(16) = −0.2,p= 0.9], suggesting that both types of TBS caused similar effects. Percent change in reaction time decreased over time in the pseudoword task [t(50) = −5.9,p< 0.001], indicating faster pseudoword processing speed with better performance 60–70 min after stimulation. In contrast, no change was demonstrated over time for the behavioral control task [t(43) = −0.6,p= 0.6], suggesting that the change over time seen in the test condition was not a learning effect. These findings provide insight into the effects of TBS on the reading system and can guide future study designs.more » « less
-
High quality Machine Learning (ML) models are often considered valuable intellectual property by companies. Model Stealing (MS) attacks allow an adversary with black-box access to a ML model to replicate its functionality by training a clone model using the predictions of the target model for different inputs. However, best available existing MS attacks fail to produce a high-accuracy clone without access to the target dataset or a representative dataset necessary to query the target model. In this paper, we show that preventing access to the target dataset is not an adequate defense to protect a model. We propose MAZE -- a data-free model stealing attack using zeroth-order gradient estimation that produces high-accuracy clones. In contrast to prior works, MAZE uses only synthetic data created using a generative model to perform MS. Our evaluation with four image classification models shows that MAZE provides a normalized clone accuracy in the range of 0.90x to 0.99x, and outperforms even the recent attacks that rely on partial data (JBDA, clone accuracy 0.13x to 0.69x) and on surrogate data (KnockoffNets, clone accuracy 0.52x to 0.97x). We also study an extension of MAZE in the partial-data setting and develop MAZE-PD, which generates synthetic data closer to the target distribution. MAZE-PD further improves the clone accuracy 0.97x to 1.0x) and reduces the query budget required for the attack by 2x-24x.more » « less
-
To enable building and testing models on long-document comprehension, we introduce QuALITY, a multiple-choice QA dataset with context passages in English that have an average length of about 5,000 tokens, much longer than typical current models can process. Unlike in prior work with passages, our questions are written and validated by contributors who have read the entire passage, rather than relying on summaries or excerpts. In addition, only half of the questions are answerable by annotators working under tight time constraints, indicating that skimming and simple search are not enough to consistently perform well. Our baseline models perform poorly on this task (55.4%) and significantly lag behind human performance (93.5%).more » « less
-
Abstract Speech processing often occurs amid competing inputs from other modalities, for example, listening to the radio while driving. We examined the extent to which dividing attention between auditory and visual modalities (bimodal divided attention) impacts neural processing of natural continuous speech from acoustic to linguistic levels of representation. We recorded electroencephalographic (EEG) responses when human participants performed a challenging primary visual task, imposing low or high cognitive load while listening to audiobook stories as a secondary task. The two dual-task conditions were contrasted with an auditory single-task condition in which participants attended to stories while ignoring visual stimuli. Behaviorally, the high load dual-task condition was associated with lower speech comprehension accuracy relative to the other two conditions. We fitted multivariate temporal response function encoding models to predict EEG responses from acoustic and linguistic speech features at different representation levels, including auditory spectrograms and information-theoretic models of sublexical-, word-form-, and sentence-level representations. Neural tracking of most acoustic and linguistic features remained unchanged with increasing dual-task load, despite unambiguous behavioral and neural evidence of the high load dual-task condition being more demanding. Compared to the auditory single-task condition, dual-task conditions selectively reduced neural tracking of only some acoustic and linguistic features, mainly at latencies >200 ms, while earlier latencies were surprisingly unaffected. These findings indicate that behavioral effects of bimodal divided attention on continuous speech processing occur not because of impaired early sensory representations but likely at later cognitive processing stages. Crossmodal attention-related mechanisms may not be uniform across different speech processing levels.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Journal Article:
- https://doi.org/10.5070/G6011190
