Follow Hulu's official WeChat account for the latest company updates and recruitment information.
This article explains how Hidden Markov Models can be used to model Chinese word segmentation, covering the underlying Markov process, model parameters, basic HMM problems, and both supervised and unsupervised training methods.
This article introduces Generative Adversarial Networks (GANs), explains their minimax formulation, value function, Jensen‑Shannon divergence, common variants, and practical training issues such as gradient saturation, while also previewing the next topic on Hidden Markov Models.
This article explains why traditional gradient descent struggles with massive datasets, introduces stochastic gradient descent and mini‑batch gradient descent as efficient alternatives, and provides practical guidance on batch size selection, data shuffling, and learning‑rate scheduling for deep learning models.
This article explains the motivation, core principles, and implementation details of Batch Normalization, including how it normalizes each mini‑batch, restores learned feature distributions, and is applied in convolutional neural networks to speed up training and boost model generalization.
This article explains the fundamentals of feature engineering for structured data, covering why and how to normalize numerical features, various categorical encoding techniques, methods for creating high‑dimensional interaction features, and decision‑tree based strategies for efficiently discovering valuable feature combinations.
This article uses realistic Hulu business scenarios to illustrate the pitfalls of relying solely on accuracy, precision, recall, RMSE, and other single metrics, and explains how combining complementary evaluation measures such as average accuracy, precision‑recall curves, ROC, F1‑score, and MAPE can provide a more comprehensive assessment of classification, ranking, and regression models.
This article revisits Principal Component Analysis by framing it as a minimum‑square‑error regression problem, showing how the optimal projection line aligns with linear regression, deriving the solution in both two‑dimensional and high‑dimensional spaces, and linking it to the classic maximum‑variance approach.
This article introduces the fundamentals of topic modeling, compares the probabilistic latent semantic analysis (pLSA) and latent Dirichlet allocation (LDA) methods, explains their graphical models and inference via EM or Gibbs sampling, and discusses practical strategies for selecting the optimal number of topics using perplexity or hierarchical Dirichlet processes.
This article explains how kernel functions and slack variables empower Support Vector Machines to achieve zero training error on linearly inseparable data, presents three theoretical questions about Gaussian kernels, error‑free classification without slack variables, and the impact of the regularization parameter C when using SMO, and provides detailed analytical solutions.
This article explains why vanilla Stochastic Gradient Descent often struggles in deep learning, describes the challenges of valleys and saddle points, and introduces three major SGD variants—Momentum, AdaGrad, and Adam—detailing their motivations, update rules, and advantages.
