New Approach for Hierarchical Classifier Training and Multi-level Image Annotation

2008 ◽  
pp. 45-57 ◽  
Author(s):  
Jianping Fan ◽  
Yuli Gao ◽  
Hangzai Luo ◽  
Shin’ichi Satoh
Keyword(s):  
Image Annotation ◽  
New Approach ◽  
Multi Level ◽  
Hierarchical Classifier ◽  
Classifier Training

NEW APPROACH TO MULTI-LEVEL PROCESSOR SCHEDULING

2010 ◽  
Vol 19 (03) ◽  
pp. 335-346 ◽  
Author(s):  
SAMANEH HOSSEINI SEMNANI ◽  
KAMRAN ZAMANIFAR
Keyword(s):  
Fitness Function ◽  
Systems Design ◽  
Heuristic Function ◽  
Problem Space ◽  
New Approach ◽  
Processor Scheduling ◽  
Quantum Time ◽  
Multi Level ◽  
Evaluation Parameters

The problem of finding the best quantum time in multi-level processor scheduling is addressed in this paper. Processor scheduling is one of the most important issues in operating systems design. Different schedulers are introduced to solve this problem. In one scheduling approach, processes are placed in different queues according to their properties, and the processor allocates time to each queue iteratively. One of the most important parameters of a processor's efficiency in this approach is the amount of time slices associated to each processor queue. In this paper, an ant colony optimization (ACO) algorithm is presented to solve the problem of finding appropriate time slices to assign to each processor queue. In this technique, each ant tries to find an appropriate scheduling. Ant algorithm searches the problem space to find the best scheduling. The quality of each ant's solution is evaluated using a new fitness function. This fitness function is designed according to the evaluation parameters of each processor queue and also according to the queue theory's relations. Also a heuristic function is presented which prompts ant to select better solutions. Computational tests are presented and the comparisons made with genetic algorithm (GA) and particle swarm optimization (PSO) algorithms which try to solve same problem. The results show the efficiency of this algorithm.

scholarly journals Automated Development of Individual Learning Paths on the Competency Approach

2013 ◽  
Vol 3 (4) ◽  
pp. 27
Author(s):  
Alexander Evgenevich Shukhman ◽  
Maria Vladimirovna Motyleva ◽  
Isabella Davidovna Belonovskaya
Keyword(s):  
Information System ◽  
Petri Net ◽  
Individual Learning ◽  
Educational Institutions ◽  
New Approach ◽  
Learning Path ◽  
Multi Level ◽  
Learning Paths ◽  
The Individual ◽  
Competency Level

This article proposes new approach to the development of student�??s individual learning path (ILP). The result of student�??s learning is defined as a set of competencies levels. The use of the multi-level system of generalized professional competencies allows to model ILP effectively and to automate the development of an optimal ILP. Each competency level in our model introduces some substantial units called subunits. The process of learning is a sequence of transitions from one student�??s state to another through a study of one or more courses. It is possible to present the relation of competency levels and courses in our model as Petri net. Using the Petri net model it is possible to automatically generate the individual learning path including the courses that the student needs to attend according to the set of certain competency levels. For searching of an optimum individual learning path we use algorithm based on tree of reachability for Petri net. The developed model and algorithm were used in the prototype of the information system for automating the construction of individual learning paths. The developed information system may be used for the construction of individual student's learning paths in educational institutions of higher education.

scholarly journals Automatic Image Annotation by Sequentially Learning From Multi-Level Semantic Neighborhoods

IEEE Access ◽  
2021 ◽  
Vol 9 ◽  
pp. 135742-135754
Author(s):  
Houjie Li ◽  
Wei Li ◽  
Hongda Zhang ◽  
Xin He ◽  
Mingxiao Zheng ◽  
...  
Keyword(s):  
Image Annotation ◽  
Automatic Image Annotation ◽  
Multi Level

scholarly journals Improving radar performance with constant envelope multi-level chirp

2019 ◽  
Vol 13 (2) ◽  
pp. 100-103
Author(s):  
Farhad Bahadori-Jahromi ◽  
Alireza Hassannejad
Keyword(s):  
Data Transmission ◽  
Additive White Gaussian Noise ◽  
Continuous Phase ◽  
Rician Fading ◽  
Telecommunication System ◽  
New Approach ◽  
Chirp Modulation ◽  
Constant Envelope ◽  
Transmission Period ◽  
Multi Level

M-level Continuous Phase Chirp Modulation is considered for data transmission. In this article, we would like to propose a new approach to transmitting information with the help of multi-level wave guide signaling. The main problem with the classic chirp modulation system is that the final phase of the signal is not equal in each period to the initial phase of the next step, and this discontinuity is present in each data transmission period. The main objective of this paper is to improve the efficiency of chirp modulation, which is recommended for continuous phase multi-level scalar modulation, which results has a significant increase in system performance. In this new approach, for sending the M symbols, we use the M level to send the message. With the help of this idea, we simulate the telecommunication system and examine the probability of its bit error in the presence of additive white Gaussian noise and Rician fading.

scholarly journals Multi-Level Head-Wise Match and Aggregation in Transformer for Textual Sequence Matching

2020 ◽  
Vol 34 (05) ◽  
pp. 9209-9216
Author(s):  
Shuohang Wang ◽  
Yunshi Lan ◽  
Yi Tay ◽  
Jing Jiang ◽  
Jingjing Liu
Keyword(s):  
Language Processing ◽  
State Of The Art ◽  
Vector Representation ◽  
Sequence Pair ◽  
Sequence Matching ◽  
New Approach ◽  
Multiple Tasks ◽  
Pair Matching ◽  
Multi Level ◽  
Multiple Levels

Transformer has been successfully applied to many natural language processing tasks. However, for textual sequence matching, simple matching between the representation of a pair of sequences might bring in unnecessary noise. In this paper, we propose a new approach to sequence pair matching with Transformer, by learning head-wise matching representations on multiple levels. Experiments show that our proposed approach can achieve new state-of-the-art performance on multiple tasks that rely only on pre-computed sequence-vector-representation, such as SNLI, MNLI-match, MNLI-mismatch, QQP, and SQuAD-binary.

Sign in / Sign up

Export Citation Format

Share Document