We deliver AI algorithm tutorials and the latest news, curated by a team of researchers from Peking University, Shanghai Jiao Tong University, Central South University, and leading AI companies such as Huawei, Kuaishou, and SenseTime. Join us in the AI alchemy—making life better!
This article introduces eight Python libraries—Open3D, Trimesh, PyVista, Vedo, Pyrender, PlotOptiX, Polyscope, and Simple‑3dviz—detailing their installation, core APIs, and step‑by‑step code examples for visualizing meshes, point clouds, and multidimensional data.
This article walks through implementing a basic multi‑layer neural network using only NumPy, covering terminology, network architecture, forward and backward propagation, activation functions, loss calculation, parameter updates with SGD, and compares the custom model with a Keras implementation.
The article presents a machine‑learning‑assisted harmonic mean estimator that computes Bayesian model evidence without dependence on sampling strategies, explains its theoretical basis, compares it to the original estimator, and demonstrates its accuracy on Rosenbrock and Normal‑Gamma benchmarks.
This article explains the fundamentals of SSH, key‑based authentication, the SSH agent workflow, how SSH agent forwarding operates, how to configure ProxyJump, and the associated security risks along with mitigation techniques.
This article explains the role of activation functions in deep learning, covering their definition, why they are needed, the main categories—including linear, binary step, and various non‑linear functions such as Sigmoid, TanH, ReLU, Leaky ReLU, ELU, Softmax and Swish—along with each function's mathematical form, advantages, disadvantages, and practical usage recommendations.
This article walks through the HTTP/1.1 keep‑alive model, shows how a simple Go program creates separate TCP connections for sequential requests, uses netstat and tcpdump to observe the behavior, and demonstrates how adjusting http.Client's transport settings enables true persistent connections even under concurrency.
The article explains why traditional planar CNNs fail on spherical data, describes how encoding rotational symmetry through continuous spherical representations and spherical harmonics leads to spherical convolutions that are rotation‑equivariant, and outlines the practical computation using harmonic coefficients.
The article introduces hybrid rotationally equivariant spherical CNNs, explains generalized signals on the sphere and rotation group, describes how linear and nonlinear operations using Clebsch‑Gordan coefficients preserve equivariance, and demonstrates efficient architectures that achieve state‑of‑the‑art results on 3D shape classification and atomic energy prediction.
This article explains how to build a Physics‑Informed Neural Network (PINN) in PyTorch to solve a simple logistic ordinary differential equation, covering the underlying theory, loss formulation with equation residuals and boundary conditions, network architecture, automatic differentiation, and training results.
The article explains how normal (Gaussian) distributions underpin many machine‑learning algorithms, reviewing the central limit theorem, multivariate Gaussian sampling, and key properties such as products, sums, conditional and marginal distributions, linear transformations, and Gaussian‑based Bayesian inference.
