Navegando por Assunto "Prediction"

Agora exibindo 1 - 2 de 2

Acesso aberto (Open Access)
Compression of activation signals from partitioned deep neural networks exploring temporal correlation
(Universidade Federal do Pará, 2024-11-27) SILVA, Lucas Damasceno; KLAUTAU JÚNIOR, Aldebaro Barreto da Rocha; http://lattes.cnpq.br/1596629769697284
The use of artificial neural networks for object detection, along with advancements in 6G and IoT research, plays an important role in applications such as drone-based monitoring of structures, search and rescue operations, and deployment on hardware platforms like FPGAs. However, a key challenge in implementing these networks on such hardware is the need to economize computational resources. Despite substantial advances in computational capacity, implementing devices with ample resources remains challenging. As a solution, techniques for partitioning and compressing neural networks, as well as compressing activation signals (or feature maps), have been developed. This work proposes a system that partitions neural network models for object detection in videos, allocating part of the network to an end device and the remainder to a cloud server. The system also compresses the feature maps generated by the last layers on the end device by exploiting temporal correlation, enabling a predictive compression scheme. This approach allows neural networks to be embedded in low-power devices while respecting the computational limits of the device, the transmission rate constraints of the communication channel between the device and server, and the network’s accuracy requirements. Experiments conducted on pre-trained neural network models show that the proposed system can significantly reduce the amount of data to be stored or transmitted by leveraging temporal correlation, facilitating the deployment of these networks on devices with limited computational power
Acesso aberto (Open Access)
Farinha de mandioca (Manihot esculenta) e tucupi: uma abordagem analítica utilizando espectroscopia no unfravermelho próximo (NIRS) e ferramentas quimiométricas
(Universidade Federal do Pará, 2022-04-25) POMPEU, Darly Rodrigues; SOUZA, Jesus Nazareno Silva de; http://lattes.cnpq.br/3640438725903079; PENA, Rosinelson da Silva; http://lattes.cnpq.br/3452623210043423
The near infrared spectroscopy (NIRS) coupled to chemometrics has been used as an alternative tool for quick and reliable solutions. Cassava flour (CF) can be classified as fermented and non-fermented types. Tucupi is a yellow broth, acidic, mostly aromatic and widely used in Regional dishes in Para state. This thesis proposed to apply for the first time the NIRS associated with chemometrics to predict quality parameters from CF and tucupi, as well as to discriminate fermented and non-fermented CF. One hundred six samples of CF was investigated and nine physicochemical parameters of CF were evaluated. Calibration equations with independent validation were developed to predict all parameters using the partial least square regression method. The performance of models was evaluated by the root mean standard error of calibration (RMSEC) and validation (RMSEV), and R2 values. The aW (RMSEC = RMSEV = 0.05), moisture content (RMSEC = 0.35%; RMSEV = 0.45%) and pH (RMSEC = 0.16; RMSEV = 0.18) could be predicted (R2 > 0.727) by NIRS coupled to multivariate analysis. NIRS coupled to Principal Component Analysis–Linear Discriminant Analysis (PCA-LDA) was also used to investigate the classification of fermented and unfermented CF. The use of NIRS spectra allows to obtain better performance parameters (training accuracy: 86.3–93.8%; validation accuracy: 84.6–96.2%) to discriminate fermented and unfermented CF than the use of the physicochemical properties (training accuracy: 80%; validation accuracy: 84.6%). NIRS was also used to predict nine quality physicochemical properties of tucupi Sixty-five samples of tucupi were used in this study. The performance of models was evaluated by the R2, RMSEC, root mean standard error of cross-validation (RMSECV) and RMSEV values. The total soluble solids contents could be predicted (R2 > 0.727; RMSEC = 0.184%; RMSECV = 0.411%; RMSEV = 0.338%) by NIRS coupled to multivariate analysis. NIRS and chemometrics proved to be a powerful tool to predict quality parameters in CF and tucupi as well as to discriminate fermented and non-fermented CF.