Multitask learning

After feature selection, there are usually many features not selected, and the information hidden in these discarded features shall be lost. In recent years, a new concept named multitask learning (MTL) has been proposed to reuse the redundant features [18 19]. The basic idea of multitask learning is to use the selected features as the input feature set, and combine the target values with some of the discarded features as the target output. It has proved that MTL can improve the accuracy of prediction of the base learning method. 

Main Task Selected Inputs Extra Features Extra Inputs Extra Outputs 

The same Extra Features when used as outputs Standard PLS using Selected Inputs as inputs and only Main Task as outputs. 

PLS using Extra Features as Extra Inputs to learn Main Task. 

---

Although the deletion of redundant features can improve the generalization ability of resultant mathematical model, in many cases, the information of the deleted features can be recovered by using these deleted features as a part of output of learning machine in mathematical modeling. This novel method is called multitask learning. This method will also be briefly described in this chapter. 

As mentioned above, three artificially generated data sets have been preprocessed by feature selection. Based on this selection, we have performed multitask learning using the partial least square method with LOO cross-validation on these three data sets. The values of errors computed for different inputs and outputs are listed in Table 4.10. 

Table 4.10 Errors of multitask learning using PLS on the three data sets.

From Table 4.10, it can be seen that the multitask learning improves the accuracy of prediction in this case. 

Multitask and Learn Creating a Curriculum Around Choosing a Curriculum... 

I have trouble multitasking. I don t learn as fast as I did. I know I come across like I m stupid, like I m an airhead. I would like to change jobs, but what if I don t remember new things ... 

By their very nature, tasks which enable one to measure spatiotemporal accuracy are complex or higher-level sensory-motor tasks. These place demands on a large number of lower-level PRs such as visual acuity, dynamic visual perception, range of movement, strength, simple reaction times, acceler-ation/deacceleration, static steadiness, dynamic steadiness, prediction, memory, open-loop movements, concentration span, attention switching, that is, central executive function or supervisory attentional system (multitask abilities), utilization of preview, and learning. 

LabVision is well adapted to design special worksheets for microreaction systems. The LabVision software includes a self-explanatory, self-documenting and multitasking programming language (HiText) for control, on-line evaluation and communication. Users can easily learn how to use it within a short time without any previous programming experience. 

---