Reducing Smartwatch Users’ Distraction with Convolutional Neural Network

Abstract

Smartwatches provide a useful feature whereby users can be directly aware of incoming notifications by vibration. However, such prompt awareness causes high distractions to users. To remedy the distraction problem, we propose an intelligent notification management for smartwatch users. e goal of our management system is not only to reduce the annoying notifications but also to provide the important notifications that users will swiftly react to. To analyze how to respond to the notifications daily, we have collected 20,353 in-the-wild notifications. Subsequently, we trained the convolutional neural network models to classify important notifications according to the users' contexts. Finally, the proposed management allows important notifications to be forwarded to a smartwatch. As experiment results show, the proposed method can reduce the number of unwanted notifications on smartwatches by up to 81%.

CASN:Class-Aware Score Network for Textual Adversarial Detection

“…Four of these methods, DISP, FGWS, RDE and MDRE have already been introduced in §2.2. We also add a detection method, MD, which simultaneously detects outof-distribution and adversarial samples (Lee et al, 2018). It first calculates the class-conditional Gaussian distribution of the features and then gives the adversarial confidence score of the samples by Mahalanobis distance.…”

A Systematic Survey on Android API Usage for Data-Driven Analytics with Smartphones

“…NotificationListenerService 31 (launched Android API level 21) and NotificationManager 32 (launched Android API level 1) can collect the more specific notification information (e.g., notification app name, time, status, priority, text, category, sound, visibility). Currently, notification related Android APIs such as NotificationListenerService and NotificationManager are widely used along with AS API to study notification usage behaviors in existing studies (e.g., [17,15,100,25,14,32]). [106].…”

“…The US API provides functions to query app usage logs by specifying time units of day, month, and year, thereby retrieving the results specified by the interval which is a more granular and explicit query system [11]. Since 2014, the US API has been mainly used to collect app usage pattern data in many studies (e.g., [12,13,14,15]). However, the AS API is still exclusively used to track the user interface (UI) status information (e.g., UI changed status, interaction type, UI properties & hierarchy, and notification) in the present studies (e.g., [16,17,18,19]).…”

“…Along with the AS/US APIs, other Android built-in sensors and APIs are also used to collect mobile usage and sensor data for the research. These data can be collected using Android built-in hardware (HW) sensors (e.g., accelerometer, gyro, GPS, Wi-Fi, Bluetooth, proximity, light, compass, pressure) and APIs (e.g., Sensor APIs 11 , Wi-Fi APIs 12 , Google Location Services APIs 13 , Bluetooth APIs 14 ). In other words, we can use these Android built-in sensors and APIs to track a user's physical activity, physiological state, and surrounding context information (e.g., location, ambient environment) for the research as shown in Figure 1.…”

What Learned Representations and Influence Functions Can Tell Us About Adversarial Examples

“…In the image domain, defence against adversarial attack can be 'proactive' or 'reactive' (Cohen et al, 2020), where proactive defence refers to improving the model's robustness (Madry et al, 2018;Gopinath et al, 2018;Cohen et al, 2019) and reactive defence focuses on detecting real adversarial examples before they are passed to neural networks (Feinman et al, 2017;Ma et al, 2018;Lee et al, 2018;Papernot and McDaniel, 2018). Broadly speaking, for reactive methods, the detection of adversarial examples involves taking a conceptualisation of the space of learned representations and the adversarial subspaces within them (Tanay and Griffin, 2016;Tramèr et al, 2017), and then characterising the differences in some function of the learned representations between the actual and the adversarial inputs produced by the DNN; for example, Ma et al (2018) applied a local intrinsic dimensionality (LID) measure to the learned representations and used that to successfully distinguish normal and adversarial images.…”