Computers use machinelearning, which is a collection of calculations and quantifiable models, to carry out anticipated tasks. Face recognition, conversation recognition, clinical determination, quantifiable exchange,...
详细信息
A powerful mechanism for detecting anomalies in a self-supervised manner was demonstrated by model training on normal data, which can then be used as a baseline for scoring anomalies. Recent studies on diffusion model...
详细信息
ISBN:
(纸本)9783031451690;9783031451706
A powerful mechanism for detecting anomalies in a self-supervised manner was demonstrated by model training on normal data, which can then be used as a baseline for scoring anomalies. Recent studies on diffusion models (DMs) have shown superiority over generative adversarial networks (GANs) and have achieved better-quality sampling over variational autoencoders (VAEs). Owing to the inherent complexity of the systems being modeled and the increased sampling times in the long sequence, DMs do not scale well to high-resolution imagery or a large amount of training data. Furthermore, in anomaly detection, DMs based on the Gaussian process do not control the target anomaly size and fail to repair the anomaly image, which led us to the development of a simplex diffusion and selective denoising ((SD)(2)) model. (SD)(2) does not require a full sequence of Markov chains in image reconstruction for anomaly detection, which reduces the time complexity, samples from the simplex noise diffusion process that have control over the anomaly size and are trained to reconstruct the selective features that help to repair the anomaly. (SD)(2) significantly outperformed the publicly available Brats2021 and Phenomena detection from X-ray image datasets compared to the self-supervised model. the source code is made publicly available at https://***/MAXNORM8650/SDSquare.
Hate speech is a threat to democratic values, because it stimulates incitement to discrimination, which international law prohibits. To limit the harmful effects of this scourge, scientists often integrate into social...
详细信息
ISBN:
(纸本)9783031631092;9783031631108
Hate speech is a threat to democratic values, because it stimulates incitement to discrimination, which international law prohibits. To limit the harmful effects of this scourge, scientists often integrate into social network platforms models provided by deep learning algorithms allowing to detect and react automatically to a message with a hateful nature. One of the particularities of these algorithms is that they are so efficient as the amount of data used is large. However, sequential execution of these algorithms on large amounts of data can take a very long time. In this paper we first compared three variants of Recurrent Neural Network (RNN) to detect hate messages. We have shown that Long Short Time Memory (LSTM) provides better metric performance, but implies more important execution time in comparison with Gated Recurrent Unit (GRU) and standard RNN. To have both good metric performance and reduced execution time, we proceeded to a parallel implementation of the training algorithms. We proposed a parallel implementation based on an implicit aggregation strategy in comparison to the existing approach which is based on a strategy with an aggregation function. the experimental results on an 8-core machine at 2.20GHz show that better results are obtained withthe parallelization strategy that we proposed. For the parallel implementation of an LSTM using the dataset obtained on kaggle, we obtained an f-measure of 0.70 and a speedup of 2.2 with our approach, compared to a f-measure of 0.65 and a speedup of 2.19 with an explicit aggregation strategy between workers.
Memory replay is crucial for learning and consolidation. Hippocampal place cells demonstrate neuronal replay of behavioral sequences at a faster timescale in forward and reverse directions during resting states (awake...
详细信息
ISBN:
(纸本)9783031451690;9783031451706
Memory replay is crucial for learning and consolidation. Hippocampal place cells demonstrate neuronal replay of behavioral sequences at a faster timescale in forward and reverse directions during resting states (awake and sleep). We propose a model of the hippocampus to demonstrate replay characteristics. the model comprises two parts - a Neural Oscillator Network to simulate replay and a Deep Value network to learn value function. the Neural Oscillator Network learns the input signal and allows modulation of the speed and direction of replay of the learned signal by modifying a single parameter. Combining reward information withthe input signal and when trained withthe Deep Value Network, reverse replay achieves faster learning of associations than forward replay in case of a rewarding sequence. the proposed model also explains the changes observed in the replay rate in an experimental study in which a rodent explores a linear track with changing reward conditions.
Embedding techniques for categorical attributes play a pivotal role in the performance of machinelearning inference, especially for downstream analysis of genomic sequence data. Categorical data do not convey quantit...
详细信息
Speaker recognition is a prominent area of study within speech technology. In recent times, there has been a notable transition towards embedding-based end-to-end speaker recognition approaches. these methods enable s...
详细信息
Tea and herbal infusions are significant to human life because they possess many medical benefits such as antioxidant and anti-inflammatory properties, among others;for that reason, the identification of the tea class...
详细信息
the proceedings contain 141 papers. the topics discussed include: mangrove species identification using deep neural network;tuning improvement of power system stabilizer using hybrid Harris Hawk optimization-equilibri...
ISBN:
(纸本)9798350399615
the proceedings contain 141 papers. the topics discussed include: mangrove species identification using deep neural network;tuning improvement of power system stabilizer using hybrid Harris Hawk optimization-equilibrium optimizer algorithm;a binarized systolic array-based neuromorphic architecture with high efficiency;context-aware recommendation system survey: recommendation when adding contextual information;monte Carlo method for map area calculation in wildland fire map management;calibration-free monocular distance estimation performance assessment under influences of environmental conditions;rice factory warehouse layout design with a combination of association rule and dedicated storage methods;torque analysis of V-type interior PMSM for electric vehicle based on FEA simulation;analysis of marketplace social media user engagement by topic;face expression recognition with local ternary pattern images using convolutional neural network and extreme learningmachine;and dual-robotic-manipulator collaborative system based on depth image.
Aiming at the problems of air separation technology, lack of research on oxygen extraction rate and insufficient generalization ability of existing prediction model of oxygen extraction rate, this paper proposes a pre...
详细信息
For ensuring global food security and sustainable agriculture, a major challenge is to control plant diseases. there is a need to improve existing procedures for early detection of plant diseases by using deep learnin...
详细信息
暂无评论