scholarly journals Understanding Dilated Mathematical Relationship between Image Features and the Convolutional Neural Network’s Learnt Parameters

Entropy ◽  
2022 ◽  
Vol 24 (1) ◽  
pp. 132
Author(s):  
Eyad Alsaghir ◽  
Xiyu Shi ◽  
Varuna De Silva ◽  
Ahmet Kondoz

Deep learning, in general, was built on input data transformation and presentation, model training with parameter tuning, and recognition of new observations using the trained model. However, this came with a high computation cost due to the extensive input database and the length of time required in training. Despite the model learning its parameters from the transformed input data, no direct research has been conducted to investigate the mathematical relationship between the transformed information (i.e., features, excitation) and the model’s learnt parameters (i.e., weights). This research aims to explore a mathematical relationship between the input excitations and the weights of a trained convolutional neural network. The objective is to investigate three aspects of this assumed feature-weight relationship: (1) the mathematical relationship between the training input images’ features and the model’s learnt parameters, (2) the mathematical relationship between the images’ features of a separate test dataset and a trained model’s learnt parameters, and (3) the mathematical relationship between the difference of training and testing images’ features and the model’s learnt parameters with a separate test dataset. The paper empirically demonstrated the existence of this mathematical relationship between the test image features and the model’s learnt weights by the ANOVA analysis.

Author(s):  
Andrii Tereshchenko ◽  
Valeriy Zadiraka

Introduction. The emergence of new parallel computational systems, such as multi-core processors, clusters, distributed systems, is due to the solution of various applied problems in various fields. The difference between devices for which parallel algorithms are implemented causes a variety of existing methods for parallelizing the calculation of multi-digit arithmetic operations. There is a problem of developing universal algorithms for implementing multi-digit arithmetic operations that are efficiently performed on various devices and on various systems. Very often it is not possible to develop a new algorithm, since at this stage there is still no test data with which it is possible to analyze the result of calculation. Therefore, the task of preparing test data and results is no less important than the development of the algorithm itself. The quality of the prepared data determines the quality of the implemented algorithm and the time required to find and eliminate errors in the algorithm-program and its implementation. In this paper, some simple dependencies are given, using which you can visually check the correctness of the calculation of multi-digit operations of addition, subtraction, multiplication, multiplication by modulo and exponentiation by modulo. Simple algorithms for generating input and output multi-digit data are presented. Using dependencies allows to check the integrity of the output when delegating computations to distributed systems such as cloud computing. The purpose of the article is to show simple dependencies between the input data and the results of performing multi-digit operations of addition, subtraction, multiplication, multiplication by modulo and exponentiation by modulo. For the given dependencies, methods for generating input and output multi-digit numbers are shown, which can be used to check the correctness of the calculation of multi-digit operations, which significantly saves the time required for preparing test data. Dependencies are provided in a generic way, which allows you to generate input data and results for devices that operate on words of different lengths (8, 16, 32, 64, 128, 256, etc. bits). Results. The dependences between the input data and the results of performing multi-digit operations are analyzed. The provided dependencies are proved in the form of lemmas. The dependencies are presented in a general form, since to generate multi-digit sequences, it is needed to set two parameters: N – the number of digits in the multi-digit value and n – the length of the digits in bits. The examples show the generation of input data and results for various multi-digit operations. Conclusions. The paper presents dependencies that are easy to remember and use for visual verification of the results of multi-digit calculations without using additional or special software or hardware, which allows to devote the saved time to developing new or more efficient modifications of multi-digit algorithms. Keywords: multi-digit arithmetic, parallel computational model.


2021 ◽  
Vol 13 (1) ◽  
pp. 152
Author(s):  
Haklim Choi ◽  
Xiong Liu ◽  
Gonzalo Gonzalez Abad ◽  
Jongjin Seo ◽  
Kwang-Mog Lee ◽  
...  

Clouds act as a major reflector that changes the amount of sunlight reflected to space. Change in radiance intensity due to the presence of clouds interrupts the retrieval of trace gas or aerosol properties from satellite data. In this paper, we developed a fast and robust algorithm, named the fast cloud retrieval algorithm, using a triplet of wavelengths (469, 477, and 485 nm) of the O2–O2 absorption band around 477 nm (CLDTO4) to derive the cloud information such as cloud top pressure (CTP) and cloud fraction (CF) for the Geostationary Environment Monitoring Spectrometer (GEMS). The novel algorithm is based on the fact that the difference in the optical path through which light passes with regard to the altitude of clouds causes a change in radiance due to the absorption of O2–O2 at the three selected wavelengths. To reduce the time required for algorithm calculations, the look-up table (LUT) method was applied. The LUT was pre-constructed for various conditions of geometry using Vectorized Linearized Discrete Ordinate Radiative Transfer (VLIDORT) to consider the polarization of the scattered light. The GEMS was launched in February 2020, but the observed data of GEMS have not yet been widely released. To evaluate the performance of the algorithm, the retrieved CTP and CF using observational data from the Global Ozone Monitoring Experiment-2 (GOME-2), which cover the spectral range of GEMS, were compared with the results of the Fast Retrieval Scheme for Clouds from the Oxygen A band (FRESCO) algorithm, which is based on the O2 A-band. There was good agreement between the results, despite small discrepancies for low clouds.


2009 ◽  
Vol 17 (1) ◽  
pp. 85-105 ◽  
Author(s):  
Walter H. Hirtle

Abstract This is an attempt to discern more clearly the underlying or POTENTIAL meaning of the simple form of the English verb, described in Hirtle 1967 as 'perfective'. Vendler's widely accepted classification of events into ACCOMPLISHMENTS, ACHIEVEMENTS, ACTIVITIES, and STATES is examined from the point of view of the time necessarily contained between the beginning and end of any event, i.e. EVENT TIME as represented by the simple form. This examination justifies the well known dynamic/stative dichotomy by showing that event time is evoked in two different ways, that, in fact, the simple form has two ACTUAL significates. Further reflection on the difference between the two types thus expressed—developmental or action-like events and non-developmental or state-like events—leads to the conclusion that the simple form provides a representation of the time required to situate all the impressions involved in the notional or lexical import of the verb.


2019 ◽  
Vol 4 (4) ◽  
pp. 421-430
Author(s):  
Gusri Akhyar Ibrahim ◽  
Arinal Hamni ◽  
Wahyu Budiono

MAKING AND TESTING OF SKEWERS CUTTING MACHINES. In Indonesia there are more than 100 types of bamboo that can be used by craftsmen, one of which is skewers. Skewers are promising commodities for business opportunities. The process of producing skewers are started from cutting down the bamboo, cutting bamboo, splitting bamboo, shriveling bamboo to become a stick skewer after that is done cutting the skewer sticks, drying sticks, polishing the sticks skewers and chopping sticks. The process of cutting a stick skewer which is done at this time is still using a simple tool, so the results are bad and the cutting time is very long. To maintain the quality of the results of a good stick skewer sticks and to increase the productivity of the stick skewers, the process of producing and testing of a skewer stick cutting machine is done. The method to produce a skewer stick cutting machine is done by designing the tool, determining the material to be used then making it. This skewer stick cutting machine is made with a press system and vertical cutting directions. From the results of testing the skewer stick cutting machine obtained that the quality of skewer cutting is good and the time required to cut is only 10 seconds. the difference is about 50 seconds faster than the hand saws used, so as to increase the productivity of the skewer sticks. The cutting machine was impelemented at home industry at Sidomulyo of South Lampung. Using the machine has increased productity and also quality of skewers.


2021 ◽  
Vol 13 (22) ◽  
pp. 4528
Author(s):  
Xin Yang ◽  
Lei Hu ◽  
Yongmei Zhang ◽  
Yunqing Li

Remote sensing image change detection (CD) is an important task in remote sensing image analysis and is essential for an accurate understanding of changes in the Earth’s surface. The technology of deep learning (DL) is becoming increasingly popular in solving CD tasks for remote sensing images. Most existing CD methods based on DL tend to use ordinary convolutional blocks to extract and compare remote sensing image features, which cannot fully extract the rich features of high-resolution (HR) remote sensing images. In addition, most of the existing methods lack robustness to pseudochange information processing. To overcome the above problems, in this article, we propose a new method, namely MRA-SNet, for CD in remote sensing images. Utilizing the UNet network as the basic network, the method uses the Siamese network to extract the features of bitemporal images in the encoder separately and perform the difference connection to better generate difference maps. Meanwhile, we replace the ordinary convolution blocks with Multi-Res blocks to extract spatial and spectral features of different scales in remote sensing images. Residual connections are used to extract additional detailed features. To better highlight the change region features and suppress the irrelevant region features, we introduced the Attention Gates module before the skip connection between the encoder and the decoder. Experimental results on a public dataset of remote sensing image CD show that our proposed method outperforms other state-of-the-art (SOTA) CD methods in terms of evaluation metrics and performance.


2017 ◽  
Vol 6 (1) ◽  
pp. 1265
Author(s):  
Nahdia Rupawanti BR ◽  
Budi Jatmiko ◽  
Soeparman Kardi

This Research aimed at developing instructional learning materials, particularly Optical Equipments, based on Learning Cycle 5E to facilitate the student’s Critical Thinking. The research’s subjects are developing instructional learning materials and students of grade XI in Islamic Senior High School Darul Istiqomah, Bojonegoro. The study used Dick’s and Carey’s model. The developed instructional learning material is tryouted using one group pretest-posttest design in the classroom. The study’s data analysis used descriptive quantitative technique and descriptive qualitative technique. The results are obtained: developing instructional learning material validity are generally categorized as valid; RPP performing is well; the dominant activity is observation; students also have positive responses to learning model; student’s critical thinking increases. The Obstacles encountered during learning is the time required to carry out learning activities exceeds from predetermined time. Based on the findings and results, it is concluded that the instructional learning materials of Learning Cycle 5E are feasible to facilitate the student’s Critical Thinking. Penelitian ini bertujuan untuk mengembangkan perangkat pembelajaran berdasarkan model Learning Cycle 5E pokok bahasan alat – alat optik untuk melatihkan kemampuan berpikir kritis siswa yang layak. Subjek penelitian adalah perangkat pembelajaran dan siswa kelas XI di MA Darul Istiqomah kabupaten Bojonegoro. Penelitian ini merupakan penelitian pengembangan dengan menggunakan model Dick and Carey. Uji coba hasil pengembangan perangkat dalam pembelajaran di kelas menggunakan one group pretest-posttest design. Teknik analisis data menggunakan deskriptif kuantitatif dan deskriptif kualititatif. Temuan dari hasil penelitian ini yaitu: Validitas perangkat pembelajaran dinyatakan valid, keterlaksanaan RPP berkategori baik; aktivitas siswa paling dominan selama pembelajaran adalah melakukan pengamatan; respon siswa terhadap pembelajaran model Learning Cycle 5E dinilai baik; kemampuan berpikir kritis siswa meningkat. Kendala yang ditemui selama pembelajaran adalah waktu yang diperlukan untuk melaksanakan kegiatan pembelajaran melebihi waktu yang telah dialokasikan. Berdasarkan pembahasan hasil dan temuan-temuan, disimpulkan bahwa perangkat pembelajaran model Learning Cycle 5E yang telah dikembangkan pada materi alat – alat optik, layak dipergunakan untuk melatihkan kemampuan berpikir kritis siswa.


Author(s):  
Hugo Luis Rojas-Villalobos ◽  
Blair Stringam ◽  
Zohrab Samani ◽  
Luis Carlos Alatorre Cejudo ◽  
Christopher Brown

Most methods for estimating the morphometric values of water bodies use equations derived from hypsographic curves or digital terrain models (DTMs) that relate depth, volume (V), and area (A) and that model the uncertainty inherent in the complex underwater morphology. This research focuses directly on the use of topobathymetric models that include the bathymetry and topography of the surrounding area next to the water body. The projection of the water surface height (H) on each DTM pixel generates a water column with intrinsic attributes such as volume and area. The process is replicated among all cells and estimates the total area and volume of the water body. If the V or A is the input data, an algorithm that iterates height values is used to generate the new data, which is compared with the entered value that functions as a reference. If the difference between the reference value and the calculated value is less than an error threshold, the iteration stops, and the maximum and average depths are calculated. The raster and the shape that represent the body of water are created. The cross comparison of H-V-A showed that there is an error between 0.0034% and 0.000039% when any of the parameters are used as input data. Performance tests determined that pixel dimensions are directly proportional to the processing time for each iteration. The results of the implementation of this algorithm were satisfactory since, for the DTM of Bustillos Lagoon, Chihuahua, Mexico, the simulation took less than 17 seconds in at most 22 iterations.


2016 ◽  
Vol 5 (1) ◽  
pp. 44 ◽  
Author(s):  
Purnama Silitonga ◽  
Mara Bangun Harahap ◽  
Derlina .

This study aims: 1) to determine differences in science process skills of students with learning model inquiry training and conventional learning models, 2) to determine the difference science process skills of students who have high creativity and creativity is low, 3) to determine the interaction model of learning inquiry trainingwith creativity of the science process skills. The sampling technique conducted cluster random sampling two classes, where first class as a class experiment with the number of students 32 people applied learning model inquiry training (X-1) and the second class as a class control the number of students 32 people who applied conventional learning model ( X-2). Instruments in this study is the science process skills test and a test of creativity in the form of a description. From these results it can be concluded that: 1) science process skills of students that learned with a learning model inquiry training is better than the students that learned with conventional learning models, 2)science process skills of students with high creativity better than students with creativity is low, 3) there is interaction between inquirylearning model training and creativity in influencing the science process skills of students.


2021 ◽  
Vol 35 (2) ◽  
pp. 105-113
Author(s):  
Wina Dhamayanti ◽  
Silvya Indraningtyas ◽  
Adji Sastrosupadi

Penelitian ini bertujuan untuk mengetahui perbedaan pengaruh antara model pembelajaran luring dan model pembelajaran daring terhadap prestasi belajar siswa SD Metta School Surabaya. Desain yang digunakan dalam penelitian berbentuk eksperimen dengan pendekatan kuantitatif. Instrumen yang digunakan dalam penelitian ini adalah angket yang disebarkan kepada 44 siswa SD dari kelas III-VI dan dokumen berupa nilai raport siswa. Data yang telah diperoleh dari penyebaran angket kemudian dianalisis menggunakan skala likert dan diuji dengan menggunakan uji-z. Hasil penelitian ini menunjukkan bahwa model pembelajaran luring dan model pembelajaran daring ada pengaruh terhadap prestasi belajar, jika ditinjau dari persen capaian sebesar 95% dan 79,75%. Selanjutnya berdasarkan uji-z diperoleh nilai signifikansi 6,827> , sehingga dapat disimpulkan bahwa ada pengaruh yang sangat nyata pada model pembelajaran luring dan model pembelajaran daring terhadap prestasi belajar. Jadi, peran guru dalam proses pembelajaran sangat penting untuk memancing motivasi belajar anak. Hal tersebut, juga harus didorong oleh usaha peserta didik dalam kegiatan belajar, sehingga pengalaman yang diperolehnya dapat memberikan manfaat dalam belajar. Kata-kata Kunci: Luring, Daring, Prestasi Belajar   Abstract: The study aims to determine the difference in the effect of offline learning model and online learning model on student achievement at Metta School Surabaya Elementary School. The design used in this research is in the form of an experiment with a quantitative approach. The instrument used in this study was a questionnaire distributed to 44 elementary school students from grades III-VI and a document in the form of student report cards. Data obtained from questionnaire distribution and then analyzed using a Likert scale and tested using the z-test. The results of this study indicate that the offline learning model and online learning model have an influence on learning achievement, when viewed from the percentage of achievement of 95% and 79,75%. Furthermore, based on the z-test, a significance value of 6,827>  was obtained, so it can be concluded that there is a very significance effect on the offline learning model and online learning model on learning achievement. Keywords: Offline Learning Model, Online Learning Model, Learning  Achievement


1968 ◽  
Vol 23 (3) ◽  
pp. 757-758 ◽  
Author(s):  
Herman S. Napier

The pooling of abilities or nominal groups technique was used in the present experiment to compare individuals with two-person groups on a picture-puzzle task. When size of the task was limited to a part (one-fourth) of the puzzle or the duration of the task was restricted to a few (four) trials, no difference between individuals and groups was evident. However, as task size and number of trials increased, groups performed at a significantly higher level than individuals. The difference was discussed in terms of information available to group members and the time required for group formation.


Sign in / Sign up

Export Citation Format

Share Document