1 views collected around this technical thread.
This tutorial walks you through building a digit classifier using PyTorch and the MNIST dataset, covering environment setup, data loading, model construction, training, evaluation, and model persistence while explaining core deep‑learning concepts.
This tutorial introduces the fundamentals of deep learning, covering neural network basics, Keras fundamentals, and provides a step‑by‑step Python example that loads the Iris dataset, preprocesses data, builds, compiles, trains, evaluates, visualizes, and predicts with a simple neural network model.
This article presents the SCNN model for multivariate time series forecasting, explains its decomposition into long‑term, seasonal, short‑term, and co‑evolving components, details the neural‑network‑based fusion and loss design, provides Python code snippets, and demonstrates its practical deployment for business volume prediction at Ctrip.
This article walks through constructing a basic neural network using only Python and NumPy, explains the underlying concepts such as neurons, training cycles, sigmoid activation, and weight‑adjustment formulas, and provides complete, runnable code with sample inputs and outputs.
This article explains the principle of spatial attention in convolutional neural networks, details the underlying algorithmic steps, and provides a complete PyTorch implementation including the attention module, full network architecture, and practical considerations for integrating spatial attention into deep learning models.
This article uses a blindfolded mountain‑climbing analogy to explain how gradient descent trains neural networks, covering cost functions, learning rates, iterative updates, and provides a Python implementation for a simple three‑layer network example.
Airbnb upgraded its neural‑network ranking system by adding a similarity network that penalizes duplicate‑like listings, enabling the algorithm to present a more diverse set of options, which boosted booking rates, value, and five‑star ratings, demonstrating that reduced result similarity improves overall search quality.
This tutorial demonstrates how to download the Fashion‑MNIST dataset, build a four‑layer fully connected neural network with PyTorch, and train it using loss functions, Adam optimizer, learning‑rate strategies, and Dropout to achieve high‑accuracy multi‑class image classification.
This paper presents a neural‑enhanced camera fingerprint framework combined with zero‑knowledge proof and digital signature schemes to reliably trace the originating device of photos, offering high‑accuracy identification, privacy preservation, and resistance to forgery across various application scenarios.
This article provides a comprehensive, step‑by‑step explanation of how ChatGPT generates text, covering token probabilities, n‑gram models, embeddings, attention mechanisms, and the Transformer architecture, while illustrating concepts with Wolfram‑language examples and visualizations.
This article demonstrates how a machine‑learning enthusiast used ChatGPT to generate, modify, and refine Keras code for training, evaluating, visualizing, and deploying a neural‑network model that classifies handwritten digits from the classic MNIST dataset, showcasing the full development workflow.
Vivo’s frontend team built a self‑attention‑based machine‑learning model that predicts web‑page layout types (column, row, or absolute) from node dimensions and positions, solving parent‑child and sibling relationships for design‑to‑code conversion, achieving 99.4% accuracy using over 20 k labeled, crawled, and generated samples, while outlining further enhancements.
The article demonstrates how to build a digital twin for lithium‑ion batteries in Python by constructing a physics‑based model, augmenting it with experimental data using a simple Keras neural network, and visualizing predictions, illustrating the hybrid approach’s improved accuracy over purely empirical methods.
This article explains how to generate low‑quality audio samples from clean speech by randomly inserting noise at various SNR levels, compute objective PESQ scores as ground‑truth, and use these paired data to train a neural‑network model for reference‑free audio quality assessment.
This article describes how to integrate and extend the open‑source HanziWriter library for Android, iOS and web, extracting TrueType glyph data, segmenting strokes with corner detection and neural‑network scoring, generating stroke medians via Voronoi diagrams, and determining correct stroke order using a Hungarian‑algorithm match.
By extracting hour‑level activity logs and training supervised models—including CART, GBDT, and neural networks—on user tags, the team simulated short‑term metrics for new reward campaigns, enabling earlier prediction of next‑day retention and shortening experiment cycles despite delayed T+1 data.
This article explains the backpropagation training algorithm for fully connected artificial neural networks, detailing its gradient‑descent basis, mathematical derivation, matrix formulation, and provides a complete Python implementation with mini‑batch stochastic gradient descent, momentum, learning‑rate decay, and experimental results.
The article details a contestant's end‑to‑end approach for an advertising competition, covering data analysis, rule‑based preprocessing, a three‑layer neural network architecture, model‑rule ensemble weighting, self‑correction strategies for the B phase, and final model‑only solutions that achieved top scores.
The article describes an intelligent cover‑image selection system for NetEase News that automatically filters unsuitable illustrations, assesses image quality with a pairwise‑trained deep model across clarity, color and composition, and smartly crops images using aspect‑ratio‑aware object detection, dramatically cutting manual editing and enabling confidence‑based automatic publishing.
Tencent’s industrial practice shows how a large‑scale offline‑nearline‑online “Shield” recommendation architecture, powered by the DeepR framework built on RCaffe, uses deep semantic embeddings, massive neural networks and reinforcement‑learning decisions to handle billions of daily requests, demonstrating that data richness and engineering capability, not model depth alone, drive performance in big‑data recommendation systems.