Dataset Information

Dynamic sampling of images from various categories for classification based incremental deep learning in fog computing.

ABSTRACT: Incremental learning evolves deep neural network knowledge over time by learning continuously from new data instead of training a model just once with all data present before the training starts. However, in incremental learning, new samples are always streaming in whereby the model to be trained needs to continuously adapt to new samples. Images are considered to be high dimensional data and thus training deep neural networks on such data is very time-consuming. Fog computing is a paradigm that uses fog devices to carry out computation near data sources to reduce the computational load on the server. Fog computing allows democracy in deep learning by enabling intelligence at the fog devices, however, one of the main challenges is the high communication costs between fog devices and the centralized servers especially in incremental learning where data samples are continuously arriving and need to be transmitted to the server for training. While working with Convolutional Neural Networks (CNN), we demonstrate a novel data sampling algorithm that discards certain training images per class before training even starts which reduces the transmission cost from the fog device to the server and the model training time while maintaining model learning performance both for static and incremental learning. Results show that our proposed method can effectively perform data sampling regardless of the model architecture, dataset, and learning settings.

SUBMITTER: Dube S

PROVIDER: S-EPMC8293927 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:Fog computing (FC) based sensor networks have emerged as a propitious archetype for next-generation wireless communication technology with caching, communication, and storage capacity services in the edge. Mobile edge computing (MEC) is a new era of digital communication and has a rising demand for intelligent devices and applications. It faces performance deterioration and quality of service (QoS) degradation problems, especially in the Internet of Things (IoT) based scenarios. Therefore, existing caching strategies need to be enhanced to augment the cache hit ratio and manage the limited storage to accelerate content deliveries. Alternatively, quantum computing (QC) appears to be a prospect of more or less every typical computing problem. The framework is basically a merger of a deep learning (DL) agent deployed at the network edge with a quantum memory module (QMM). Firstly, the DL agent prioritizes caching contents via self organizing maps (SOMs) algorithm, and secondly, the prioritized contents are stored in QMM using a Two-Level Spin Quantum Phenomenon (TLSQP). After selecting the most appropriate lattice map (32 × 32) in 750,000 iterations using SOMs, the data points below the dark blue region are mapped onto the data frame to get the videos. These videos are considered a high priority for trending according to the input parameters provided in the dataset. Similarly, the light-blue color region is also mapped to get medium-prioritized content. After the SOMs algorithm's training, the topographic error (TE) value together with quantization error (QE) value (i.e., 0.0000235) plotted the most appropriate map after 750,000 iterations. In addition, the power of QC is due to the inherent quantum parallelism (QP) associated with the superposition and entanglement principles. A quantum computer taking "n" qubits that can be stored and execute 2 n presumable combinations of qubits simultaneously reduces the utilization of resources compared to conventional computing. It can be analyzed that the cache hit ratio will be improved by ranking the content, removing redundant and least important content, storing the content having high and medium prioritization using QP efficiently, and delivering precise results. The experiments for content prioritization are conducted using Google Colab, and IBM's Quantum Experience is considered to simulate the quantum phenomena.

Dataset Information

Dynamic sampling of images from various categories for classification based incremental deep learning in fog computing.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets