Lida Li

ICML Conference 2021 Conference Paper

SimAM: A Simple, Parameter-Free Attention Module for Convolutional Neural Networks

• Lingxiao Yang
• Ru-Yuan Zhang
• Lida Li
• Xiaohua Xie

In this paper, we propose a conceptually simple but very effective attention module for Convolutional Neural Networks (ConvNets). In contrast to existing channel-wise and spatial-wise attention modules, our module instead infers 3-D attention weights for the feature map in a layer without adding parameters to the original networks. Specifically, we base on some well-known neuroscience theories and propose to optimize an energy function to find the importance of each neuron. We further derive a fast closed-form solution for the energy function, and show that the solution can be implemented in less than ten lines of code. Another advantage of the module is that most of the operators are selected based on the solution to the defined energy function, avoiding too many efforts for structure tuning. Quantitative evaluations on various visual tasks demonstrate that the proposed module is flexible and effective to improve the representation ability of many ConvNets. Our code is available at Pytorch-SimAM.

IJCAI Conference 2007 Conference Paper

• Jianqiang Shen
• Lida Li
• Thomas G. Dietterich

Desktop users commonly work on multiple tasks. The TaskTracer system provides a convenient, low-cost way for such users to define a hierarchy of tasks and to associate resources with those tasks. With this information, TaskTracer then supports the multi-tasking user by configuring the computer for the current task. To do this, it must detect when the user switches the task and identify the user's current task at all times. This problem of "task switch detection" is a special case of the general problem of change-point detection. It involves monitoring the behavior of the user and predicting in real time when the user moves from one task to another. We present a framework that analyzes a sequence of observations to detect task switches. First, a classifier is trained discriminatively to predict the current task based only on features extracted from the window in focus. Second, multiple single-window predictions (specifically, the class probability estimates) are combined to obtain more reliable predictions. This paper studies three such combination methods: (a) simple voting, (b) a likelihood ratio test that assesses the variability of the task probabilities over the sequence of windows, and (c) application of the Viterbi algorithm under an assumed task transition cost model. Experimental results show that all three methods improve over the single-window predictions and that the Viterbi approach gives the best results.

AAAI Conference 2005 System Paper

The TaskTracker System

• Simone Stumpf
• Anton Dragunov
• Jon Herlocker
• Lida Li

Knowledge workers spend the majority of their working hours processing and manipulating information. These users face continual costs as they switch between tasks to retrieve and create information. The TaskTracer project at Oregon State University investigates the possibilities of a desktop software system that will record in detail how knowledge workers complete tasks, and intelligently leverage that information to increase efficiency and productivity. Our approach assigns each observed user interface action to a task for which it is likely being performed. In this demonstration we show how we have applied machine learning in this environment.