Wiki

Learning With Errors (LWE) is a very important problem in cryptography and theoretical computer science, proposed by Oded Regev in 2005. The LWE problem can be described as: given a system of linear equations, where each […]

In mathematics, low-rank approximation is a minimization problem where the cost function measures the goodness of fit between a given matrix (the data) and an approximation matrix (the optimization variables), but the rank of the approximation matrix must be reduced.

Knowledge distillation is a machine learning technique that aims to transfer the learnings of a large pre-trained model (the “teacher model”) to a smaller “student model”.

YOLOv10 achieves state-of-the-art performance while significantly reducing computational overhead

Infrastructure as a Service (IaaS) is a cloud computing service that provides the necessary computing, storage, and network resources on a pay-as-you-go basis.

NAS refers to storage devices that connect to a network and provide file access services to computer systems.

Data lakes are different from data warehouses or silos in that they use a flat architecture with object storage to maintain metadata for files.

The General Data Protection Regulation (GDPR) is the strictest privacy and security law in the world.

Hyper Converged Infrastructure (HCI) combines servers and storage into a distributed infrastructure platform, creates flexible building blocks through intelligent software, and replaces traditional infrastructure consisting of separate servers, storage networks, and storage arrays.

Exascale computing refers to computing systems capable of computing at least “10 18 IEEE 754 double-precision (64-bit) operations (multiplications and/or additions) per second (exa FLOPS)” and is a standard measure of supercomputer performance. Exascale computing is computing […]

HyperNetworks is a neural network structure that has some differences in model parameterization compared to traditional neural networks. The paper "HyperNetworks" published by Google Brain in 2016 stated that in HyperNetworks, the model parameterization of the model is different from that of traditional neural networks.

Predictive Coding (PC) is a theoretical framework in cognitive science that holds that the human brain processes cognition through spatiotemporal predictions of the visual world.

The diffusion probability model demonstrates the connection between the diffusion probability model and PC theory.

The DQ-LoRe framework utilizes Dual Query (DQ) and Low-Rank Approximate Reranking (LoRe) to automatically select contextual learning examples.

Contrastive learning is a technique that enhances the performance of vision tasks by using the principle of contrasting samples against each other to learn properties that are common between data classes and properties that distinguish one class from another.

By addressing the limitations of traditional LSTM and incorporating novel components such as exponential gating, matrix memory, and parallelizable architecture, xLSTM opens up new possibilities for LLM.

A point cloud is a dataset of points in space that can represent a 3D shape or object, typically acquired by a 3D scanner.

Referring Image Segmentation (RIS) aims to segment the target object referred to by natural language expressions. However, previous methods rely on a strong assumption that a sentence must describe an object in an image.

The multiple drafts model is a physicalist theory of consciousness based on cognitivism, proposed by Daniel Dennett. The theory views the mind from the perspective of information processing. Dennett published Consciousness Explained in 1991.

KAN: Kolmogorov-Arnold Networks The paper proposes a promising alternative to Multilayer Perceptron (MLP) called Kolmogorov-Arnold Networks (KAN). The name KAN comes from […]

The Kolmogorov-Arnold representation theorem makes it easier to analyze complex dynamical systems

Action model learning encompasses a complex process within the field of artificial intelligence where models are developed essentially to predict the effects of an agent’s actions in an environment.

True Positive Rate (TPR) is a metric used in statistics, machine learning, and medical diagnosis to evaluate the performance of binary classification models. It represents the proportion of actual positive cases that are correctly identified or classified as positive by the model. TPR is also known as sensitivity, recall, or […]

Glitch tokens are words that, in large language models, are supposed to help the model run smoothly, but result in abnormal output. A research team from Huazhong University of Science and Technology, Nanyang Technological University, and other universities published a study in 2024 titled "Glitch Tokens in […]