Query-Based Learning System
(Redirected from Query-Based Learning)
Jump to navigation
Jump to search
A Query-Based Learning System is a Active Learning System in which a student (learner) learns a target concept by having a dialogues with a teacher.
- AKA: Query-Based Learning Mechanism System, Learning With Queries System.
- Context
- It can solve a Query-Based Learning Task by implementing a Query-Based Learning Algorithm.
- It can range from being an Unsupervised Query-Based Learning System to being a Supervised Query-Based Learning System.
- Example(s):
- Counter-Example(s):
- See: Machine Learning System, PAC Learning, Reinforment Learning, Deep Learning.
References
2018
- (Chang et al., 2018) ⇒ Ray-I Chang, Chien-Chang Huang, and Chia-Yun Lee (2018). "Query-Based Machine Learning Model for Data Analysis of Infrasonic Signals in Wireless Sensor Networks". In: Proceedings of the 2nd International Conference on Digital Signal Processing. DOI:10.1145/3193025.3193031
- QUOTE: Big data increases the computation complexity, and the wrong selection of features may decreases the accuracy in event prediction. To overcome this problem, a query-based-learning method is applied to select the proper features for smart edge computing in machine learning. Experimental results show that the proposed method provides good performance when comparing with previous feature selection methods.
2017
- (Jain & Stephan, 2017) ⇒ Sanjay Jain, Frank Stephan. (2017). “Query-Based Learning”. In: (Sammut & Webb, 2017). DOI: 10.1007/978-1-4899-7687-1_694
- QUOTE: Query learning models the learning process as a dialogue between a pupil (learner) and a teacher; the learner has to figure out the target concept by asking questions of certain types and whenever the teacher answers these questions correctly, the learner has to learn within the given complexity bounds. Complexity can be measured by both, the number of queries as well as the computational complexity of the learner. Query learning has close connections to statistical models like PAC learning (...) Most learning scenarios consider learning as a relatively passive process where the learner observes more and more data and eventually formulates a hypothesis that explains the data observed. Query-based learning is an active learning process where the learner has a dialogue with a teacher, which provides on request useful information about the concept to be learned.
2011
- (Sencer, 2011) ⇒ Safiye Sencer (2011). "Query Based Learning in Multi-Agent Systems".
- QUOTE: Query based learning is a part of machine learning which optimizes the performance criterion using current and past situation data. The model defined up to some parameters, and learning is the execution and optimizes the parameters of the model using the past experience. The [suggested model able descriptive to gain knowledge from data, derivative to obtain rule from knowledge base with query then predict future decisions. When interface realizes the query process, consider the following requirements. In order to identify and use the characteristics relevant to the task to be taken into outline in the interface level components are: information and resources; control parameters and activations. Information and resources includes detail information about task, cases, attributes with name, goal, index and hierarchy framework. The dynamic knowledge base presents with input and output interface. Control parameters aim to check and detect some failures in the system with temporal constraints, error toleration and functional limits. During the realization of the query based learning, system realizes some activation such as querying, reasoning, collaborating, planning and acting (Fig. 2).
Figure 2. Interaction of learning activity.=== 2005 ===
- QUOTE: Query based learning is a part of machine learning which optimizes the performance criterion using current and past situation data. The model defined up to some parameters, and learning is the execution and optimizes the parameters of the model using the past experience. The [suggested model able descriptive to gain knowledge from data, derivative to obtain rule from knowledge base with query then predict future decisions. When interface realizes the query process, consider the following requirements. In order to identify and use the characteristics relevant to the task to be taken into outline in the interface level components are: information and resources; control parameters and activations. Information and resources includes detail information about task, cases, attributes with name, goal, index and hierarchy framework. The dynamic knowledge base presents with input and output interface. Control parameters aim to check and detect some failures in the system with temporal constraints, error toleration and functional limits. During the realization of the query based learning, system realizes some activation such as querying, reasoning, collaborating, planning and acting (Fig. 2).
- (Ge et al., 2005) ⇒ Esther Ge, Richi Nayak, and Yuefeng Li (2005). "The user query based learning system for lifetime prediction of metallic components".
- QUOTE: We present the User Query Based Learning System (UQBLS) in Figure 1. It includes all phases as explained necessary in the industry standard data mining process model CRISP-DM (Cross Industry Standard Process for Data Mining) [10]. We highlight three procedures that are different from the CRISP-DM. They are critical for the success of our UQBLS model. The feature selection based on a user query is separated from data preprocessing. An external domain knowledge base is involved in data preprocessing and results post-processing phases.
Fig. 1. User Query Based Learning System.=== 2004 ===
- QUOTE: We present the User Query Based Learning System (UQBLS) in Figure 1. It includes all phases as explained necessary in the industry standard data mining process model CRISP-DM (Cross Industry Standard Process for Data Mining) [10]. We highlight three procedures that are different from the CRISP-DM. They are critical for the success of our UQBLS model. The feature selection based on a user query is separated from data preprocessing. An external domain knowledge base is involved in data preprocessing and results post-processing phases.
- (Angluin, 2004) ⇒ Dana Angluin (2004). "Queries revisited". Theoretical Computer Science, 313(2). DOI:10.1016/j.tcs.2003.11.004
- QUOTE: Formal models of learning reflect a variety of differences in tasks, sources of information, prior knowledge and capabilities of the learner, and criteria of successful performance. In the model of exact identification with queries [1], the task is to identify an unknown concept drawn from a known concept class using queries to gather information about the unknown concept.
1997
- (Chang & Hsiao, 1997) ⇒ Ray-I Chang, and Pei-Yung Hsiao (1997). "Unsupervised query-based learning of neural networks using selective-attention and self-regulation". In: IEEE transactions on neural networks, 8(2), 205-217. DOI:10.1109/72.557657
- QUOTE: Query-based learning (QBL) has been introduced for training a supervised network model with additional queried samples. Experiments demonstrated that the classification accuracy is further increased. Although QBL has been successfully applied to supervised neural networks, it is not suitable for unsupervised learning models without external supervisors. In this paper, an unsupervised QBL (UQBL) algorithm using selective-attention and self-regulation is proposed. Applying the selective-attention, we can ask the network to respond to its goal-directed behavior with self-focus
1988
- (Angluin, 1988) ⇒ Dana Angluin (1988). "Queries and concept learning". Machine learning, 2(4), 319-342. DOI:10.1007/BF00116828
- QUOTE: A successful learning component in an expert system will probably rely heavily on queries to its instructors. For example, Sammut and Bancrji's (1986) system uses queries about specific examples as part of its strategy for efficiently learning a target concept. Shapiro‘s (19817 19827 1983) Algorithmic Debugging System uses a variety of types of queries to the user to pinpoint errors in Prolog programs. In this paper, we use a formal framework to study the power of several types of queries for concept-learning tasks.
We consider the problem of identifying an unknown set [math]\displaystyle{ L }[/math] from some finite or countable hypothesis space [math]\displaystyle{ L_1, L_2, \cdots }[/math] of subsets of a universal set [math]\displaystyle{ U }[/math]. The usual assumption is that one is given an arbitrarily or stochastically generated sequence of elements of [math]\displaystyle{ U }[/math] each classified as to whether it is in [math]\displaystyle{ L_* }[/math][1]. Instead, we will assume that the learning algorithm has access to a fixed set of oracles that will answer specific kinds of queries about the unknown concept [math]\displaystyle{ L_* }[/math].
- QUOTE: A successful learning component in an expert system will probably rely heavily on queries to its instructors. For example, Sammut and Bancrji's (1986) system uses queries about specific examples as part of its strategy for efficiently learning a target concept. Shapiro‘s (19817 19827 1983) Algorithmic Debugging System uses a variety of types of queries to the user to pinpoint errors in Prolog programs. In this paper, we use a formal framework to study the power of several types of queries for concept-learning tasks.
- ↑ . We direct the reader to Angluin and Smith (1983) for a survey of inductive inference.