RELATIUM

A research space focused on intelligent systems, computational theory and cognitive concepts.

From Information Retrieval to Information Reconstruction

After multiple development stagnations in the 1960s, 1980s, and 2007, AI entered a technology acceleration phase in the post-pandemic era. The pandemic drove global digitalization, rapidly expanding infrastructure and computing power, further promoting AI application breakthroughs. In this context, AI gradually shifted from research-driven to practical application, having differential impacts on different groups. On one hand, users can improve information processing and work efficiency through AI. On the other hand, professionals also face uncertainty and competitive pressure from technological intervention.

Phenomenon Analysis

1/5/2025

K-Nearest Neighbors Algorithm Theory

K-Nearest Neighbors (KNN) is a non-parametric, instance-based supervised learning method widely applied to classification and regression problems. Its core idea is not building explicit models, but performing local inference on unknown samples through distance relationships between samples. This article systematically organizes KNN from mathematical perspectives, methodological processes, and practical applications.

Machine Learning

10/5/2025

Decision Tree Classifier Experiment

This decision tree model is based on the ID3 (Iterative Dichotomiser 3) algorithm, implemented from scratch to fully understand data splitting, impurity calculation, feature selection for splitting, and tree structure generation. This approach avoids the black-box problems of off-the-shelf packages, allowing model behavior and experimental results to be clearly tracked and analyzed.

Experiment

11/15/2025

Convolutional Neural Networks (CNN)

Convolutional Neural Network is a feedforward neural network specialized for processing grid-like data such as images, audio, and time series. This note focuses on CNN's basic principles and key components with TensorFlow/Keras implementation examples. CNN's core design inspiration comes from biological visual cortex, effectively extracting spatial hierarchical features through local connectivity and weight sharing mechanisms. Typical CNN architecture includes convolutional layers, pooling layers, and fully connected layers.

Deep Learning

3/29/2025

Cognitive Concepts

Artificial Intelligence (AI)

Artificial Intelligence is a cross-disciplinary concept whose core lies not in single definitions but continuous exploration of whether machines can possess intelligence. With technological development and application expansion, AI gradually shifts from engineering problems to philosophical and cognitive dimensions. This article organizes basic understanding of AI across philosophical, conceptual, and theoretical levels as preliminary concept documentation. AI understanding shows multiple divergent perspectives from industry technology routes, academic authority definitions, and philosophical ontology/everyday experience approaches, causing semantic fluidity of the term 'AI.'

Machine Learning

8/9/2024

New Era Engineering Landscape: Software Technical Debt Risk and Response

GenAI is like post-industrial-era automation machinery arms, indeed replacing repetitive work in software engineering and other fields. While software engineering is complex, many common program logics remain, and despite complexity, they are legacy technology and repetitive development, exactly what GenAI excels at, given its training on existing technical articles and knowledge. This unavoidably gives large enterprises opportunities for restructuring and workforce adjustment, a harsh but realistic situation. However, when GenAI's generation becomes unreliable...

Phenomenon Analysis

7/10/2025

Seventy Years of Paradigm Shifts in Artificial Intelligence History

The history of Artificial Intelligence is not a single-line accumulation of technology, but a seventy-year methodological competition. Starting from the Dartmouth Workshop in 1956, AI gradually divided into two core paths: Symbolism and Connectionism. The former believes intelligence can be decomposed into logical rules and symbolic reasoning, while the latter argues intelligence should emerge from systems similar to neural neurons in the brain.

Phenomenon Analysis

1/20/2026

Latest Articles

瀏覽更多