Vector-Based Computer Learning Using Target Processing Nodes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current computer learning methods require significant human and material resources, are inefficient, and often fail to produce effective programs for complex tasks, especially in areas like Go, where artificially programmed programs lag behind human capabilities until machine learning is applied.
Innovation Solution
A method and apparatus for computer learning that converts data into vectors, determines a target processing node from a set of processing nodes, and processes the data using this node to achieve a processing result, employing Boolean functions and logical operations to create candidate and target functions based on training data pairs, thereby improving data processing capability without relying on artificial presuppositions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional artificial programming is used to develop computer programs, then programs can be created with clear understanding of computing theories, but huge human and material resources are required and development efficiency is low
Solution Approach 1:
The system enables computers to automatically learn and generate programs through machine learning algorithms, eliminating the need for human programmers to manually write code. The learning model autonomously processes training data pairs and generates target processing nodes without human intervention,实现ing self-service program development
Solution Approach 2:
The patent replaces the mechanical process of manual programming with an automated machine learning system. Instead of human programmers mechanically writing code based on computing theories, the system uses learning models that automatically process data and generate programs through algorithms like gradient descent and backpropagation
2Reliability
If traditional artificial programming is used for complex tasks like Go, then programs can be developed with established computing theories, but effective programs are difficult to obtain and playing ability remains far below human players
Solution Approach 1:
The system changes the fundamental parameters of program development by transitioning from rule-based programming to parameter-based machine learning. The learning model adjusts parameters such as weights and biases in neural networks to optimize performance on complex tasks like Go, enabling the program to achieve playing ability comparable to human experts
Solution Approach 2:
The patent introduces dynamic learning capabilities that allow the system to adapt and improve over time. The learning model continuously processes training data and updates its internal parameters, enabling it to dynamically adjust to complex tasks and achieve reliable performance on tasks like Go that were previously intractable
3Reliability
If machine learning technologies like deep learning are introduced to improve program ability, then playing ability can reach unprecedented levels, but the learning is not all-purpose and requires significant resources
Solution Approach 1:
The patent creates a universal learning model that can handle multiple types of tasks through a single unified framework. The system processes various data types including images, text, and sensor data using the same learning architecture, enabling one model to perform multiple functions and making the learning approach all-purpose rather than task-specific
Solution Approach 2:
The system uses vector representations to create simplified copies of complex data. By converting diverse data types into vector formats, the learning model can process different kinds of information through the same mechanisms, enabling versatile application across different domains while maintaining efficient resource utilization
Data Source
AI summary
A method and device for learning by a computer, and a non-transitory computer readable storage medium, relating to the technical field of computers. The method includes: transforming data to be processed into a vector to be processed (110); determining a corresponding target processing node (120); and processing said vector (130).