scholarly journals MIDI2vec: Learning MIDI embeddings for reliable prediction of symbolic music metadata

Semantic Web ◽  
2021 ◽  
pp. 1-21
Author(s):  
Pasquale Lisena ◽  
Albert Meroño-Peñuela ◽  
Raphaël Troncy
Keyword(s):  
Digital Libraries ◽  
Feed Forward Neural Network ◽  
Latent Factors ◽  
Reliable Prediction ◽  
New Approach ◽  
Real World Applications ◽  
Low Dimensionality ◽  
Limited Scale ◽  
Comparable Accuracy ◽  
Time Signature

An important problem in large symbolic music collections is the low availability of high-quality metadata, which is essential for various information retrieval tasks. Traditionally, systems have addressed this by relying either on costly human annotations or on rule-based systems at a limited scale. Recently, embedding strategies have been exploited for representing latent factors in graphs of connected nodes. In this work, we propose MIDI2vec, a new approach for representing MIDI files as vectors based on graph embedding techniques. Our strategy consists of representing the MIDI data as a graph, including the information about tempo, time signature, programs and notes. Next, we run and optimise node2vec for generating embeddings using random walks in the graph. We demonstrate that the resulting vectors can successfully be employed for predicting the musical genre and other metadata such as the composer, the instrument or the movement. In particular, we conduct experiments using those vectors as input to a Feed-Forward Neural Network and we report good comparable accuracy scores in the prediction with respect to other approaches relying purely on symbolic music, avoiding feature engineering and producing highly scalable and reusable models with low dimensionality. Our proposal has real-world applications in automated metadata tagging for symbolic music, for example in digital libraries for musicology, datasets for machine learning, and knowledge graph completion.

SENTENCE ALIGNMENT USING FEED FORWARD NEURAL NETWORK

2006 ◽  
Vol 16 (06) ◽  
pp. 423-434 ◽  
Author(s):  
MOHAMED ABDEL FATTAH ◽  
FUJI REN ◽  
SHINGO KUROIWA
Keyword(s):  
Neural Network ◽  
Language Processing ◽  
Training Data ◽  
Parameter Vector ◽  
Feed Forward Neural Network ◽  
New Approach ◽  
Feed Forward ◽  
Parallel Corpora ◽  
Feature Parameter ◽  
Essential Resource

Parallel corpora have become an essential resource for work in multi lingual natural language processing. However, sentence aligned parallel corpora are more efficient than non-aligned parallel corpora for cross language information retrieval and machine translation applications. In this paper, we present a new approach to align sentences in bilingual parallel corpora based on feed forward neural network classifier. A feature parameter vector is extracted from the text pair under consideration. This vector contains text features such as length, punctuate score, and cognate score values. A set of manually prepared training data has been assigned to train the feed forward neural network. Another set of data was used for testing. Using this new approach, we could achieve an error reduction of 60% over length based approach when applied on English–Arabic parallel documents. Moreover this new approach is valid for any language pair and it is quite flexible approach since the feature parameter vector may contain more/less or different features than that we used in our system such as lexical match feature.

scholarly journals COLLABORATIVE FILTERING: A NEW APPROACH TO SEARCHING DIGITAL LIBRARIES

2004 ◽  
Vol 10 (2) ◽  
pp. 177-191 ◽  
Author(s):  
Janet Webster ◽  
Seikyung Jung ◽  
Jon Herlocker
Keyword(s):  
Collaborative Filtering ◽  
Digital Libraries ◽  
New Approach

Macroanalysis

2013 ◽  
Author(s):  
Matthew L. Jockers
Keyword(s):  
Digital Libraries ◽  
Large Scale ◽  
Computational Analysis ◽  
Literary Interpretation ◽  
Digital Form ◽  
New Approach ◽  
Demographic Groups ◽  
Linguistic Patterns ◽  
Digital Material ◽  
The Individual

This book introduces readers to large-scale literary computing and the revolutionary potential of macroanalysis—a new approach to the study of the literary record designed for probing the digital-textual world as it exists today, in digital form and in large quantities. Using computational analysis to retrieve key words, phrases, and linguistic patterns across thousands of texts in digital libraries, researchers can draw conclusions based on quantifiable evidence regarding how literary trends are employed over time, across periods, within regions, or within demographic groups, as well as how cultural, historical, and societal linkages may bind individual authors, texts, and genres into an aggregate literary culture. Moving beyond the limitations of literary interpretation based on the “close-reading” of individual works, the book describes how this new method of studying large collections of digital material can help us to better understand and contextualize the individual works within those collections.

KLL±approximate quantile sketches over dynamic datasets

2021 ◽  
Vol 14 (7) ◽  
pp. 1215-1227
Author(s):  
Fuheng Zhao ◽  
Sujaya Maiyya ◽  
Ryan Wiener ◽  
Divyakant Agrawal ◽  
Amr El Abbadi
Keyword(s):  
Approximation Algorithm ◽  
Data Stream ◽  
Small Space ◽  
Data Set ◽  
Real World Applications ◽  
Comparable Accuracy ◽  
Data Elements ◽  
Quantile Approximation ◽  
Optimal Construction ◽  
Standing Problem

Recently the long standing problem of optimal construction of quantile sketches was resolved byKarnin,Lang, andLiberty using the KLL sketch (FOCS 2016). The algorithm for KLL is restricted to online insert operations and no delete operations. For many real-world applications, it is necessary to support delete operations. When the data set is updated dynamically, i.e., when data elements are inserted and deleted, the quantile sketch should reflect the changes. In this paper, we proposeKLL±, the first quantile approximation algorithm to operate in thebounded deletionmodel to account for both inserts and deletes in a given data stream. KLL±extends the functionality of KLL sketches to support arbitrary updates with small space overhead. The space bound for KLL±is [EQUATION], where ∈ and δ are constants that determine precision and failure probability, and α bounds the number of deletions with respect to insert operations. The experimental evaluation of KLL±highlights that with minimal space overhead, KLL±achieves comparable accuracy in quantile approximation to KLL.

scholarly journals Surface-Controlled Photocatalysis and Chemical Sensing of TiO2, α-Fe2O3, and Cu2O Nanocrystals

Crystals ◽  
2019 ◽  
Vol 9 (3) ◽  
pp. 163 ◽  
Author(s):  
Anna Kusior ◽  
Milena Synowiec ◽  
Katarzyna Zakrzewska ◽  
Marta Radecka
Keyword(s):  
Gas Sensing ◽  
Synthesis And Characterization ◽  
Abrupt Increase ◽  
Electrical And Optical Properties ◽  
Oxide Semiconductors ◽  
New Approach ◽  
Low Dimensionality ◽  
Shape Controlled ◽  
Gas Molecules

A relatively new approach to the design of photocatalytic and gas sensing materials is to use the shape-controlled nanocrystals with well-defined facets exposed to light or gas molecules. An abrupt increase in a number of papers on the synthesis and characterization of metal oxide semiconductors such as a TiO2, α-Fe2O3, Cu2O of low-dimensionality, applied to surface-controlled photocatalysis and gas sensing, has been recently observed. The aim of this paper is to review the work performed in this field of research. Here, the focus is on the mechanism and processes that affect the growth of nanocrystals, their morphological, electrical, and optical properties and finally their photocatalytic as well as gas sensing performance.

A New Finite Fracture Mechanics Approach for Assessing the Strength of Bonded Lap Joints

2011 ◽  
Vol 471-472 ◽  
pp. 1075-1080 ◽  
Author(s):  
Philipp Weißgraeber ◽  
Wilfried Becker
Keyword(s):  
Fracture Mechanics ◽  
Energy Release ◽  
Optimization Problem ◽  
Strength Criterion ◽  
Lap Joints ◽  
Reliable Prediction ◽  
New Approach ◽  
Single Lap Joints ◽  
Finite Fracture Mechanics ◽  
Failure Loads

For the widespread use of adhesive joints an exact and reliable prediction of the strength is mandatory. In this work, a new approach to assess the strength of single lap joints is presented. The approach is based on the hybrid criterion as postulated by Leguillon in the framework of finite fracture mechanics. It strictly combines a consideration of an energy release balance and a fulfillment of a strength criterion. The present work is based on a simple model of the joint behavior and assumptions about crack initiation. From the stress distribution of the classical shear lag theory an incremental energy release rate is derived and is used to formulate the optimization problem of the failure load. The resulting predictions of critical failure loads are compared to experimental results of single lap joints. It is shown that the new approach is able to physically describe crack formation and the corresponding critical load within the framework and limitations of the underlying assumptions and simplifications. The work closes with a discussion of the limitations and an outlook on possible improvements of the underlying models and assumptions.

scholarly journals Time-Coordinated Signatures

2017 ◽  
Author(s):  
Santi J. Vives
Keyword(s):  
Real World ◽  
The State ◽  
New Approach ◽  
Real World Applications

Hash-based signatures are typically stateful: they need to keep a state with the number of past signatures to know which values have been already used and cannot be reused. If the memory storing the state fails, the security would degrade. Some implementations solve the problem by using a number of secret values so large that the probability of picking the same at random is negligible, but this solution can make the signatures impractical for some real world applications. This paper proposes a new approach to hash-based signatures: we show that it is possible to derive their state entirely from time, without the need to keep a state with the number of past signatures,

scholarly journals A Constrained Multi-Objective Learning Algorithm for Feed-Forward Neural Network Classifiers

2017 ◽  
Vol 7 (3) ◽  
pp. 1685-1693
Author(s):  
M. Njah ◽  
R. El Hamdi
Keyword(s):  
Neural Network ◽  
Learning Algorithm ◽  
Binary Classification ◽  
Benchmark Problems ◽  
Feed Forward Neural Network ◽  
New Approach ◽  
Feed Forward ◽  
Multi Objective ◽  
Adaptive Penalty ◽  
Neural Network Classifiers

This paper proposes a new approach to address the optimal design of a Feed-forward Neural Network (FNN) based classifier. The originality of the proposed methodology, called CMOA, lie in the use of a new constraint handling technique based on a self-adaptive penalty procedure in order to direct the entire search effort towards finding only Pareto optimal solutions that are acceptable. Neurons and connections of the FNN Classifier are dynamically built during the learning process. The approach includes differential evolution to create new individuals and then keeps only the non-dominated ones as the basis for the next generation. The designed FNN Classifier is applied to six binary classification benchmark problems, obtained from the UCI repository, and results indicated the advantages of the proposed approach over other existing multi-objective evolutionary neural networks classifiers reported recently in the literature.

Nanoscopic Molecular Engineering: Organic MBE and STM

1992 ◽  
Vol 247 ◽  
Author(s):  
Masahiko Hara ◽  
Anthony F. Garito ◽  
Hiroyuki Sasabe
Keyword(s):  
Molecular Beam ◽  
Molecular Engineering ◽  
Scanning Tunneling ◽  
Tunneling Microscopy ◽  
New Approach ◽  
Molecular Systems ◽  
Novel Material ◽  
Low Dimensionality ◽  
Low Dimensional ◽  
Low Dimensional Materials

ABSTRACTApplication of molecular beam epitaxy (MBE) and scanning tunneling microscopy (STM), especially for organic molecular systems, has been drawing our attention as a new approach to realizing novel material structures with low dimensionality, which are expected to exhibit important electronic and photonic properties. The following is an outline of our work in progress, including an overview of “nanoscopic” molecular engineering. Future possibilities of fabrication, modification and characterization for organic low-dimensional materials are reviewed with the new concept embraced by “nanoscopic” science and technology.

Feature Extraction of Robot Sensor Data Using Factor Analysis for Behavior Learning

2004 ◽  
Vol 8 (3) ◽  
pp. 284-294
Author(s):  
Wai-keung Fung ◽  
◽  
Yun-hui Liu
Keyword(s):  
Factor Analysis ◽  
Feature Extraction ◽  
Sensor Data ◽  
Range Data ◽  
Perceptual Space ◽  
Latent Factors ◽  
New Approach ◽  
Robot Behavior ◽  
Behavior Learning ◽  
Sensor Information

The paper addresses feature extraction of sensor data for robot behavior learning using factor analysis. Redundancies in sensor types and quantities are common in sensing competence of robots. The redundancies cause the high dimensionality of the perceptual space. It is impractical to incorporate all available sensor information in decision-making and learning of robots due to the huge memory and computational requirements. This paper proposes a new approach to extract important knowledge from sensor data based on the inter-correlation of sensor data using factor analysis and construct logical perceptual space for robot behavior learning. The logical perceptual space is constructed by hypothetical latent factors extracted using factor analysis. Since the latent factors extracted have fewer dimensions than raw sensor data, using the logical perceptual space in behavior learning would significantly simplify the learning process and architecture. Experiments have been conducted to demonstrate the process of logical perceptual space extraction from ultrasonic range data for robot behavior learning.

Sign in / Sign up

Export Citation Format

Share Document