A Practical Guide to Qdrant for RAG Applications
Published:
A practical overview of A Practical Guide to Qdrant for RAG Applications, covering A Practical Guide to Qdrant for RAG Applications, Why a vector database is needed in RAG…
Posts by Category
ai-engineering
cv
Fully Convolutional Networks
Published:
A summary of FCN (Fully Convolutional Networks) Fully Convolutional Networks for Semantic Segmentation GitHub 传统的卷积神经网络（CNN）通常包含卷积层和全连接层。在这篇论文中，作者认识到全连接层会导致输出固定大小的特征向量，从而限制了CNN在…
Convolutional Neural Networks
Published:
什么是卷积神经网络（CNN）？它在计算机视觉中的应用是什么？ 卷积神经网络（Convolutional Neural Network，CNN）是一种深度学习模型，特别适用于处理具有网格结构的数据，如图像和视频。 CNN的核心思想是通过卷积操作和池化操作来提取输入数据的特征，并通过这些特征进行分类、识别或回归等任务。 基本原理 ref
Donut: an OCR-free Document Understanding Transformer
Published:
A practical overview of Donut: an OCR-free Document Understanding Transformer, covering Encoder, Decoder, Output Conversion.
Object Detection Algorithms
Published:
A practical overview of Object Detection Algorithms, covering Important notes, Loss Function, R-CNN.
evaluation
RAG Evaluation in Practice: What to Measure and Why It Matters
Published:
A good answer does not necessarily mean a good system. Learn how to evaluate retrieval, generation, and end-to-end performance in RAG applications, and why evaluation is often a diagnosis problem rather than a scoring problem.
information-retrieval
Reranking in RAG: Why Retrieval Is Not Enough
Published:
“Reranking is a critical but often overlooked component in RAG systems. While retrieval maximizes recall, reranking improves precision by re-evaluating candidates with more expressive models. This post explains why reranking is necessary and how it fits into modern RAG pipelines.”
Retrieval in RAG: From Vector Search to Hybrid Systems
Published:
A practical introduction to retrieval in RAG systems, covering dense vector search, sparse lexical retrieval, hybrid retrieval, and the trade-offs behind real-world retrieval design.
ml
Boruta Feature Selection
Published:
A practical overview of Boruta Feature Selection, covering Purpose, How It Works, Advantages.
Clustering alrogithms
Published:
A practical overview of Clustering alrogithms, covering Improvements for gradient descent, Momentum (惯性保持), AdaGrad (环境感知).
Activation Functions
Published:
A practical overview of Activation Functions, covering Sigmoid Family, Hard Sigmoid, Swish.
Loss Functions
Published:
A practical overview of Loss Functions, covering Regression Loss Functions, L1 Loss (Mean Absolute Error, MSE), L2 Loss (Mean Square Error, MSE).
Classification Algorithms
Published:
A practical overview of Classification Algorithms, covering Assumption, Logistic Regression, Description.
Time-series Forecasting
Published:
A practical overview of Time-series Forecasting, covering ARIMA (Autoregressive Integrated Moving Average), Description, Strengths.
Metrics to Evaluate Predictive Models
Published:
A practical overview of Metrics to Evaluate Predictive Models, covering MAE (Mean Absolute Error), Definition, Formula.
personal
Happy 31th birthday
Published:
今天不谈机器学习，不再整理一些记不住的知识，而是在三十又一的当下，记录一下最近的思考。 为什么最近会有时间做这么多思考呢？因为我在今年1月，农历新年前不到1个月的时候，试用期还有1周就转正的时候，突然被通知不转正了。 到现在应该失业正好差不多两个月，中间农历新年期间完全是一个招聘停滞的状态。然后在我31岁生日之前至少有一个比较过得去的口头offer，也…
python
Python try / except / else / finally learning notes
Published:
A practical overview of Python try / except / else / finally learning notes, covering 核心概念, 基本结构, except 优先捕捉能处理的特定异常类型.
rag
A Practical Guide to Qdrant for RAG Applications
Published:
A practical overview of A Practical Guide to Qdrant for RAG Applications, covering A Practical Guide to Qdrant for RAG Applications, Why a vector database is needed in RAG…
LLM/RAG Learning notes
Published:
A practical overview of LLM/RAG Learning notes, covering 主要内容, LangChain主线总结报告.
rag-systems
RAG Evaluation in Practice: What to Measure and Why It Matters
Published:
A good answer does not necessarily mean a good system. Learn how to evaluate retrieval, generation, and end-to-end performance in RAG applications, and why evaluation is often a diagnosis problem rather than a scoring problem.
Reranking in RAG: Why Retrieval Is Not Enough
Published:
“Reranking is a critical but often overlooked component in RAG systems. While retrieval maximizes recall, reranking improves precision by re-evaluating candidates with more expressive models. This post explains why reranking is necessary and how it fits into modern RAG pipelines.”
Retrieval in RAG: From Vector Search to Hybrid Systems
Published:
A practical introduction to retrieval in RAG systems, covering dense vector search, sparse lexical retrieval, hybrid retrieval, and the trade-offs behind real-world retrieval design.
Embedding in RAG: Why Representation Matters More Than You Think
Published:
Introduction
reliability
What Makes a RAG System Reliable?
Published:
Why do RAG systems still fail even when retrieval works? This article examines grounding, trust, and common failure modes, and discusses what it takes to build reliable RAG applications.
stats
Kernel Density Estimation
Published:
A practical overview of Kernel Density Estimation, covering Kernel Density Estimation, Nonparametric Estimation, The univariate case (1 variable).