Computer-assisted Music Composition Algorithm Design Dependent on Interactive Genetic Algorithm with Interval Fitness

Computer-assisted music composition refers to computer-assisted music composition with the participation of people. However, there are problems such as style and expression. In this paper, a computer-assisted music composition algorithm based on the interactive genetic algorithm with interval fitness is proposed. A new music prediction model is established by integrating melody units and rhythms into traditional models with only notes or rhythms as units. Moreover, the generated music phrases are optimized by the interactive genetic algorithmphrase. The simulation results suggest that the proposed algorithm can generate music phrases quickly with a certain melody logic that conforms to the personal demand of users using a small data set.

Using Psychologically-Informed Priors for Suicide Prediction in the CLPsych 2021 Shared Task

This paper describes our approach to theCLPsych 2021 Shared Task, in which weaimed to predict suicide attempts based onTwitter feed data. We addressed this challengeby emphasizing reliance on prior domainknowledge. We engineered novel theory drivenfeatures, and integrated prior knowledgewith empirical evidence in a principledmanner using Bayesian modeling. Whilethis theory-guided approach increases bias andlowers accuracy on the training set, it was successfulin preventing over-fitting. The modelsprovided reasonable classification accuracy onunseen test data (0.68 ≤ AUC ≤ 0.84). Ourapproach may be particularly useful in predictiontasks trained on a relatively small data set.

Learning Local Descriptor for Comparing Renders with Real Images

We present a method to train a deep-network-based feature descriptor to calculate discriminative local descriptions from renders and corresponding real images with similar geometry. We are interested in using such descriptors for automatic industrial visual inspection whereby the inspection camera has been coarsely localized with respect to a relatively large mechanical assembly and presence of certain components needs to be checked compared to the reference computer-aided design model (CAD). We aim to perform the task by comparing the real inspection image with the render of textureless 3D CAD using the learned descriptors. The descriptor was trained to capture geometric features while staying invariant to image domain. Patch pairs for training the descriptor were extracted in a semisupervised manner from a small data set of 100 pairs of real images and corresponding renders that were manually finely registered starting from a relatively coarse localization of the inspection camera. Due to the small size of the training data set, the descriptor network was initialized with weights from classification training on ImageNet. A two-step training is proposed for addressing the problem of domain adaptation. The first, “bootstrapping”, is a classification training to obtain good initial weights for second training step, triplet-loss training, that provides weights for extracting the discriminative features comparable using l2 distance. The descriptor was tested for comparing renders and real images through two approaches: finding local correspondences between the images through nearest neighbor matching and transforming the images into Bag of Visual Words (BoVW) histograms. We observed that learning a robust cross-domain descriptor is feasible, even with a small data set, and such features might be of interest for CAD-based inspection of mechanical assemblies, and related applications such as tracking or finely registered augmented reality. To the best of our knowledge, this is the first work that reports learning local descriptors for comparing renders with real inspection images.

Modeling for exponential growth and decay methodology in biometry using SAS syntax

This paper provided an alternative method for exponential growth modeling as a regression analysis technique through the SAS algorithm. This alternative method is a combination technique (using nonlinear model bootstrap and fuzzy regression) for the small data set and gives the researcher an option to start the analysis, even if there is not enough data set. This method enhances the previous methodology with embedded bootstrapping and fuzzy technique to a nonlinear regression model. This principle aims to propose an alternative method of analysis with better results. In our case, we applied this principle to farm data and compared the results obtained by looking at the average width of the predicted interval.

Intonational variation in third party complaints in Spanish

Very little previous research has addressed the prosodic characteristics of third party complaints. This paper discusses an utterance and word level intonational analysis of this speech act in four speakers of Mexican Spanish. The effect of social distance/ power relationships was incorporated into the study by creating an experimental data elicitation task in which participants addressed identical complaints to a friend, as well as a boss, based on a series of hypothetical contexts. Major global findings revealed that all speakers significantly increased their fundamental frequency (F0) mean when directing their complaints to friends, however, only two speakers significantly expanded their F0 range in the same circumstance. Locally, peaks and valleys were manifested at significantly higher levels across the board when addressing friends. Finally, while speakers produced complaint contours of similar overall shape regardless of hearer, individual variation was present in the form of circumflex versus suppressed utterance-final F0 configurations. Overall, the relatively small data set initiated preliminary thoughts on the application of both cross-linguistic and language- and dialect-specific intonational concepts to complaints while also emphasizing the importance of relationships between interlocutors for future studies.