Bitcoin Price Prediction Using Machine Learning Algorithms

Advances in engineering research/Advances in Engineering Research, Journal Year: 2023, Volume and Issue: unknown, P. 389 - 396

Published: Jan. 1, 2023

Language: Английский

---

Music mode analysis and teaching enlightenment research under the background of digital education

Soft Computing, Journal Year: 2023, Volume and Issue: unknown

Published: June 17, 2023

Language: Английский

---

Deformer: Denoising Transformer for Improved Audio Music Genre Classification

Applied Sciences, Journal Year: 2023, Volume and Issue: 13(23), P. 12673 - 12673

Published: Nov. 25, 2023

Audio music genre classification is performed to categorize audio into various genres. Traditional approaches based on convolutional recurrent neural networks do not consider long temporal information, and their sequential structures result in longer training times convergence difficulties. To overcome these problems, a traditional transformer-based approach was introduced. However, this employs pre-training momentum contrast (MoCo), technique that increases computational costs owing its reliance extracting many negative samples use of highly sensitive hyperparameters. Consequently, complicates the process risk learning imbalances between positive sample sets. In paper, method for called Deformer proposed. The learns deep representations data through denoising process, eliminating need MoCo additional hyperparameters, thus reducing costs. it prior decoder reconstruct patches, thereby enhancing interpretability representations. By calculating mean squared error loss reconstructed real can learn more refined representation data. performance proposed experimentally compared with two distinct baseline models: one S3T employing residual network-bidirectional gated unit (ResNet-BiGRU). achieved an 84.5% accuracy, surpassing both ResNet-BiGRU-based (81%) S3T-based (81.1%) models, highlighting superior classification.

Language: Английский

---

Music Recommendation System Using Deep Learning and Machine Learning

Published: Feb. 9, 2024

This article makes use of a test dataset music to systems connected between clients and recommend new track them based on their past usage. Similarity measures the Count Vectorizer have also been used. Through this, flask front side will display suggested whenever particular song is digested. The importance managing looking for songs has increased along with quick development digital formats. Despite success Music Information Retrieval (MIR) frameworks throughout last couple years, soundtrack content - recommendation evolution remains in its beginning stages. As result, this investigates broad framework cutting-edge methods. It was discovered that two popular optimization techniques summarization information perform well. Because difficult long-tail discovery process efficacious dramatic tension soundtrack, relevant user methodologies concept sound prototype gained foothold. paper provides insights into three critical components classification method: client sculpting, item segmentation, suit algorithms. Four potential problems relating experience are explained six models. subjective suggestion method hasn't thoroughly studied, though. In order do we provide motivation-based model empirical research psychology music, sports education, human behavior. Our novel recommender system convolutional neural network (CNN) recurrent (CRNN) combination. uses deep learning analyses complex audio features tailored recommendations, improving field discovery. Using CNN, acquired average 0.724 using CRNN 0.748.

Language: Английский

---

Musical Genres Classification Utilizing the Pre-trained ResNet50 CNN Model and Deep Learning Techniques

Published: Feb. 21, 2024

This research investigates the application of ResNet50 Convolutional Neural Network (CNN) within a model framework for purpose classifying musical genres. The objective is to enhance accuracy and efficiency automated music genre categorization systems through utilization deep learning techniques. proposed employs methodology that processes raw audio data, involving extraction relevant innovative features convolutional layers. These layers are designed capture hierarchical patterns inherent specific incorporation architecture in machine facilitates temporal relationships, allowing recognize subtle nuances variations compositions. study utilizes diverse dataset encompassing multiple genres robustness adaptability model. primary goal validate effectiveness CNN Model accurately Through rigorous experimentation assessment, this aims contribute significantly advancement analysis classification systems. findings have noteworthy implications various applications, including recommendation systems, content tagging, streaming services.

Language: Английский

---