Abstract
With the exponential rise of the number of IoT devices, the amount of data being produced is massive. Thus, it is unfeasible to send all the raw data directly to the cloud for processing, especially for data that is high dimensional. Training deep learning models incrementally evolves the model over time and eliminates the need to statically training the models with all the data. However, the integration of class incremental learning and the Internet of Things (IoT) is a new concept and is not yet mature. In the context of IoT and deep learning, the transmission cost of data in the edge-cloud architecture is a challenge. We demonstrate a novel sample selection method that discards certain training images on the IoT edge device that reduces transmission cost and still maintains class incremental learning performance. It can be unfeasible to transmit all parameters of a trained model back to the IoT edge device. Therefore, we propose an algorithm to find only the useful parameters of a trained model in an efficient way to reduce the transmission cost from the cloud to the edge devices. Results show that our proposed methods can effectively perform class-incremental learning in an edge-cloud setting.
Highlights
In the computer vision domain, deep learning has shown a great amount of success and in some tasks even surpassing the level of human accuracy
This paper largely extends the work of [16] whereby data sampling is performed at the Internet of Things (IoT) edge device, in [16], the new data samples do not belong to novel classes
As learning a different number of classes at a time means learning a different number of samples at a time, we would like to observe the performance of our Discard Counting (DDC) algorithm when it learns a different number of samples at a time
Summary
In the computer vision domain, deep learning has shown a great amount of success and in some tasks even surpassing the level of human accuracy. A lot of this success has been obtained in an offline setting whereby all the data is already present on a machine before training starts and deep learning models are trained on big datasets just once and they are deployed. Even if all the data is available, it is challenging to train deep learning models as it requires powerful hardware to train such models and it is time-consuming to train on a huge amount of data altogether. Another problem with training a deep learning model offline is that once it is trained and deployed, the model will not learn any more parameters in the future. Deep learning models should be able to learn in a continuous environment
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call