Soo-Mook Moon's research while affiliated with Seoul National University and other places

... Therefore, in order to evaluate Rule 2 efficiently, we restrict it to a concrete case where two dependencies must be triggered by the same value, such as T 1 and T 3 in Example 1 (dependency is triggered by the same value A). Such a situation is common in the blockchain because, in general, smart contract access to accounts is always skewed [41,42], namely, some accounts are frequently accessed, also known as active accounts. ...

... In the literature, some recent works exploit MEC for computational offloading of DNN tasks under a single WD setup [11]- [21] or a multi-WD setup [21], [22]. The DNN task offloading and resource allocation schemes are designed to optimize the WD's energy consumption [12]- [14], the DNN inferencing accuracy [15], [16], and the DNN inferencing time [17]- [20]. The profiling knowledge of layer-wise DNN inferencing delay/energy consumption, which heavily depends on the MEC system parameters, is determined in either an offline manner [12]- [18] or by an online learning approach [19], [20]. ...

... Hybrid Collaboration combines at least two of three ECCI application patterns above or integrates additional CI/EI capabilities into ECCI applications. For example, ShadowTutor [8] enables robust HD video semantic segmentation with significant throughput improvement and network data transmission reduction. Here, cloud servers conduct both the inference of the heavy and general 'teacher' model and the training of the lightweight 'student' model. ...

... This has led to a new trend of AR using on-screen smartphone instructions [54,55]. AR remote assistance, such as "see what I see" apps, can provide better knowledge transfer via a peer-to-peer connection that incorporates video, audio, or hand annotations [56][57][58]. Other studies of AR assistance have provided remote technical support, including those for maintenance [52], manufacturing [54,59], automotive [60], and utilities [61], with visual instructions or virtual user manuals [53,55] by means of annotations and content uploaded to phones, tablets, and AR glasses. ...

... The Web Worker API 4 has been introduced to enable parallel JavaScript programming on the web. Many prior studies have shown that offloading computation-intensive tasks in web applications from mobile devices to edge or cloud servers can greatly enhance device performance without requiring the devices to have advanced compute capabilities or high connection bandwidth [3,6,8]. This was done by offloading the computation-intensive tasks to worker threads (or web workers for short), which are spawned to run concurrently with the main thread in JavaScript, and dispatching the web workers to remote servers. ...

... IONN [3] uses the shortest path method and the penalty factor method to determine DNN partitions, and builds the DNN model incrementally when each DNN partition arrives, allowing the client to start partial uploading even before uploading the entire DNN model, thereby improving query performance. Enhanced Partitioning [32] is based on a penalty factor method of uploading overhead and uses the shortest path method on the DNN execution graph between the client and the cloud/edge server to partition the DNN layer, which generates a more granular uploading plan. JointDNN [33] transforms the optimal computing scheduling problem of DNN into the shortest path problem and Integer Linear Programming (ILP) in the mobile cloud computing environment, and divides the DNN architecture by optimization formulations at layer granularity, thereby achieving collaborative computing between mobile devices and the cloud. ...

... JavaScript classification is a challenging problem due to the dynamic non-deterministic behavior of the code [67], especially when attempted on-the-fly using low-end mobile phones, since this leaves no room for executing the JS codewhich tends to consume numerous resources and time. Existing research on JS filtering primarily focuses on the identification of advertising and/or tracking JS [42,47,49], in addition to malicious JS code snippets [32,33,44,66,68,69]. ...