About haotianblog
haotianblog is a bilingual technical blog maintained by Zhonghaotian. It focuses on practical questions in machine learning, C and Python algorithms, network protocols, AI security, and battery modeling. The aim is to turn a technical claim into inputs, steps, outputs, and limitations that a reader can inspect.
Most articles begin with a small runnable problem: why backtracking finds all 92 eight-queens solutions, why K-means++ is more stable than random centroids, which TLS 1.3 flights contribute to latency, or how to prevent leakage when EIS features enter a model. Equations, code, figures, and downloads are organized around the question rather than added as decoration.
Author and maintainer
Zhonghaotian is the primary author and maintainer of the articles, code explanations, and experiment materials on this site. The writing process uses compilers, Python environments, browser developer tools, protocol clients, and plotting tools to check important steps. When an example uses synthetic data or a simplified teaching model, the article distinguishes it from production data.
A program running once is not treated as sufficient evidence. Before publication, the useful questions are whether the input is explicit, randomness is reproducible, metrics are interpretable, failure conditions are preserved, and another reader can independently check the main result.
Writing and verification method
| Stage | Material kept in the article | How a reader can check it |
|---|---|---|
| Define the problem | Inputs, outputs, assumptions, and excluded scope | Confirm that the example matches the intended use case |
| Build a baseline | A minimal equation, naive implementation, or unoptimized result | Compare the later change against the baseline |
| Run the experiment | Commands, parameters, data split, output files, and metrics | Reproduce at least one key result in the stated environment |
| Test failure cases | Boundary inputs, numerical errors, misconfiguration, and failure modes | Change one condition and observe whether the claim still holds |
| State limitations | Teaching simplifications, synthetic data, performance, and safety boundaries | Decide what additional validation is needed for another project |
Current topic tracks
- Machine learning and deep learning: begin with the machine-learning workflow, feature engineering, and evaluation, then continue to backpropagation, convolution, attention, and small classifiers.
- Algorithms in C and Python: use eight-queens backtracking, bitmask optimization, and the Iris K-means implementation to study state representation, complexity, and experiment audits.
- Network protocols: move from DNS caching, CIDR, TCP congestion windows, and TLS 1.3 to proxies, CDNs, and cache revalidation.
- AI security: connect threat modeling, adversarial examples, poisoning, model privacy, and RAG or agent tool boundaries through defensive experiments.
- Battery modeling: document PyBaMM, EIS features, aging simulation, SOH/RUL labels, and training-data audit methods.
Code, data, and figures
Downloads are not standalone attachments. Code and datasets are linked from an explanatory article where practical, with versions, input formats, commands, expected outputs, and known limitations. Network and security material defaults to local or controlled labs. Battery and machine-learning articles distinguish measured, public, and synthetic data so that simulation is not presented as a production result.
Standards, official documentation, and third-party tools provide protocol or API context. The site’s own contribution is the derivation, implementation tradeoffs, run records, and failure analysis around those sources.
Corrections and contact
Dependencies, browsers, and cloud services change. Articles are revised when a reproduction fails, an interface changes, or an explanation is incomplete. Substantial revisions add commands, results, or limitations instead of silently replacing the conclusion.
To report a problem with an equation, code sample, link, translation pair, or download, use the contact page or email haotianblog@gmail.com. Include the page URL, environment, and observed output when possible.