Example Sentences:

• Yong Zhuang: The DL models start with the collection of most comprehensive and potentially relevant datasets available for decision making process. The DL scenarios are designed to meet some performance goals to select the most appropriate DL architecture after training the model using the labeled data. The iterative training process optimizes different learning parameters, which will be ‘tweaked’ until the network provides a satisfactory level of performance.source

• Chengjie Zheng: This idea is supported by experiments that demonstrate that features learned by distinguishing time transformations capture video dynamics more than supervised learning and that such features generalize well to classic vision tasks such as action recognition or time-related task such as video synchronization. source

Before & After:

• Yong Zhuang

  • before: Space and time are omnipresent in measurements in many fields, including climate science, criminology, and earth science. Many data collection methods are designed to record each measurement’s spatial and temporal information in the data because of the Spatio-temporal nature of the real-world processes studied in these fields. Massive Spatio-temporal data provides more possibilities for research in these fields.

  • after: Across the sciences, including climatology, criminology, and earth science, researchers always use measurements distributed in space and time to understand the systems they study. Because of the Spatio-temporal nature of the real-world processes studied in these fields, many data collection methods are designed to record each measurement’s spatial and temporal information in the data. The growing size of these Spatio-temporal data sets has opened up the potential for pattern discovery and insight in these fields.

• Chengjie Zheng

  • before: Since video-level comparative learning sees information from the entire video, it helps to capture the global context. However, the execution inherent sequence structure is weak.

  • after: Video-level contrastive learning helps to capture global context as it sees information from the entire video, but it is weak in enforcing the inherent sequential structure.