Why does the hierarchical network model have certain inference ability?

Answer]: b, c, d

A characteristic feature of the hierarchical network model is its cognitive economic principles. In a network structure, the properties that a class of concepts have in common are not stored with each class member, but with their subordinate concepts. The category size effect supports this view.

The researchers asked the subjects to judge the truth of the sentence, and when the subject and predicate of the sentence were at the same level, the response was the fastest. When the subject and predicate of a sentence are separated by a node, the response time is slower; When the subject and predicate are separated by two nodes, the response time is the slower. Firstly, this model only studies the extraction of information leakage from the hierarchical level of the line, ignoring the strength of the connection, that is, ignoring the role of familiarity. Second, the model ignores the typical bonding of class members, and considers that all members of the category are identical. In fact, the typical values of each member are different, for example, the concept of apple is much higher in the category of fruit than the concept of grapefruit.

Thirdly, the hierarchical network model cannot provide a reasonable explanation for negative judgments. According to the hierarchical network model, making a negative judgment usually requires a long search. But Glass et al. 1975) found that sentences that make negative judgments about sentences such as "all birds are dogs" or "bats are birds" are fast.

The theory of the hierarchical network model that briefly describes the conceptual structure is as follows:

Psychologists believe that everyone has a long-term memory system with a huge capacity, and we organize our knowledge of the world into this system through various network models. These models are a rich network of countless concepts and relationships. They are able to capture not only physical objects such as airplanes, computers, and houses, but also many abstract concepts such as family, communication, and ability.

Classical network models include the hierarchical network model and the activation diffusion model. The hierarchical network model is the first semantic memory model in cognitive psychology proposed by Quillian, which was originally proposed for computer simulations of speech comprehension and was later used to illustrate the structure of concepts.

In this theory, concepts are stored in a network of concepts in the form of nodes, and each concept has certain characteristics, which are actually concepts. The various genus concepts are organized together according to the logical subordinate relationship, and their generic relationship is represented by connecting lines, so that the concepts that have a generic relationship with each other form a conceptual network.

In the network, the higher the level of the concept, the higher the level of abstract generalization. The activation diffusion model is also a semantic memory model. But unlike the hierarchical network model, it abandons the hierarchical structure of concepts, and instead organizes concepts with semantic connections or similarities.

Difference 1: The composition is different

1. The hierarchical model organizes data into one-to-many relationships.

Structures, hierarchies employ keywords to access each part of each of these levels.

2. The mesh model uses connection instructions or pointers to determine the explicit connection relationship between data, which is a many-to-many type of data organization.

3. The relational model organizes data in the form of record groups or data tables, so as to facilitate the storage and transformation of the relationship between various geographic entities and attributes, without stratification and without pointers, which is a very effective data organization method to establish the relationship between spatial data and attribute data.

Difference 2: The advantages of the model are different

1. Advantages of hierarchical model: the data structure is relatively simple and clear, the query efficiency of the database is high, and it provides good integrity support.

2. Advantages of network model: it can describe the real world more directly, such as a node can have multiple parents, and a node can have a variety of connections; It has good performance and high access efficiency.

3. Advantages of the relational model: it is based on strict mathematical concepts; The concept is singular, and both entities and the connections between entities are represented by relationships. The retrieval and updating structure of the data is also relational; Its access path is transparent to the user, so it has higher independence, better security and confidentiality, and simplifies the work of programmers to develop and build databases.

Difference 3: The shortcomings of the model are different

1. Disadvantages of hierarchical model: many connections in the real world are non-hierarchical, and it is not suitable for many-to-many connections between nodes; The inquiry of the child's node must be through the parent's node; Due to the tight structure, hierarchical commands tend to be procedural.

2. Disadvantages of mesh model: the structure is relatively complex, and with the expansion of the application environment, the structure of the database becomes more and more complex, which is not conducive to the grasp of end users; The DDL and DML of the mesh model are complex, and they need to be embedded in a certain high-level language (C, Cobol), which is not easy for users to grasp and use.

3. Disadvantages of the relational model: The concealment of the access path leads to the query efficiency being inferior to the formatted data model.

The content described by the data model consists of three parts: data structure, data manipulation, and data constraints.

Data models are divided into three types according to different application levels: conceptual data models, logical data models, and physical data models.

Among the conceptual data models, there are the object-oriented models and predicate models you mentioned, as well as the ER model.

Hierarchical, reticular, and relational models are three important data models.

The data model corresponding to the tree diagram is a hierarchical model; The data model that corresponds to a mesh diagram is a mesh model. A relational model is an unformatted structure that uses the structure of a single two-dimensional table to represent entities and the connections between entities. A two-dimensional table that satisfies certain conditions is called a relationship.

Classification of Neural Network ModelsThere are many models of artificial neural networks, which can be classified according to different methods. Among them, the two common classification methods are according to the topological structure of the network connection and according to the information flow direction within the network. 1 Classify the topology of the network according to the topology of the network, that is, the way neurons are connected to each other.

Based on this, neural network structures can be divided into two main categories: hierarchical and interconnected. Hierarchical neural networks divide neurons into output layers, middle layers (hidden layers), and output layers according to their functions and sequences.

Each neuron in the output layer is responsible for receiving input information from the outside world and transmitting it to the neurons in the middle layer. The hidden layer is the internal information processing layer of the neural network and is responsible for information transformation. It can be designed as one or more layers as needed; The last hidden layer transmits information to the output layer, and the neurons are further processed and then outputs the information processing results to the outside world. In the interconnected network structure, there may be a connection path between any two nodes, so the interconnected network can be subdivided into three situations according to the degree of connection of the nodes in the network

Fully interconnected, partially interconnected, and sparsely connected2 According to the classification of network information flow, from the perspective of the direction of information transmission within the neural network, it can be divided into two types: feedforward network and feedback network. The structure of a simple feedforward network is the same as that of a hierarchical network, and the feedforward is named because the direction of network information processing is carried out layer by layer from the input layer to the hidden layer and then to the output layer.

In a feedforward network, the output of the previous layer is the input of the next layer, and the processing of information has the directionality of layer-by-layer transmission, and there is generally no feedback loop. As a result, such networks are easily connected in series to create multi-layer feedforward networks. The structure of a feedback network is the same as that of a single-layer fully interconnected fabric network.

All nodes in a feedback network have information processing capabilities, and each node can both receive input from and output to the outside world.

1. Hierarchical model.

A set of basic hierarchical connections that satisfies the following two conditions: (1) there is only one node and no parental node (this node is called the root node); (2) Nodes other than the root node have one or only one parental node.

The hierarchical model is similar to the mesh model in that records and links are used to represent data and the connections between data, respectively. Unlike the mesh model, records in a hierarchical model can only be organized into a collection of trees, not any graphs.

Hierarchical models can be seen as a special case of mesh models, and they are both formatted models. They all share common characteristics, from architecture to database language to data storage management. In the hierarchical model, the organization of records is no longer a disorganized diagram, but a tree"Inverted long"of the tree.

2. Mesh model.

A set of basic hierarchical connections that satisfies the following two conditions: (1) allowing more than one node to have no parental nodes; (2) A node can have multiple parental nodes.

The data in the mesh model is represented by a collection of records (which have the same meaning as records in the Pascal language), and the connections between the data are represented by links (which can be thought of as pointers). Records in a database can be organized into a collection of arbitrary graphs.

3. Relational model.

A relational model uses a collection of tables to represent data and the connections between data. Each table has multiple columns, each with a unique column name. In the relational model, both the entities abstracted from objective things and the connections between entities are represented by a single structural type - relationships.

After all the processing of the relationship, the result is the relationship - a new two-dimensional table. Follow-up: May I ask your uncle.