TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

Ask HN: Simple projects to implement machine learning

11 点作者 fjellfras将近 13 年前
I have been working through Tom Mitchell's book on machine learning and I want to supplement the theory with a practical project.<p>Does anyone have experience with small projects which helped them in a similar situation? The most I have come up with is building a news classifier which will use the feedzilla and NY Times api but if anyone has other good ideas please let me know.<p>I have reasonable programming experience with python and have access to a linode so I will be using these to implement the project.<p>Thank you

5 条评论

asit将近 13 年前
I used lua to create a simple AI. This AI can learn words and store them in a dictionary. It can link one word with another, having some resemblance or relationship. Also I implemented a "emotional link" between them; one word can be either in a GOOD or BAD relationship with another. Like the word "rat" will have a BAD relationship with "cat". It is based on meaning of words, rather than acting like a mere processor of vocabulary crawler! This AI can communicate with me.. asks me questions that it generates from meaningful concepts. Ofcourse, it is very basic right now and makes errors. But it will grow. The bottomline is that you must implement "from the roots". Perception and recognition are built over these roots. Hope you got on idea what I am attempting to build. One thing most required right now is adding a solid grammer engine to it. If I were not a lazy programmer, I would have done it already. :D
评论 #4094138 未加载
jfaucett将近 13 年前
I built an image interpretation app once that was really interesting from a machine learning perspective. Using image magick I extracted shapes and based on the layout I was able to feed that into a "composition" algorithm that judged whether the image layout was good or not. I also implemented a "plant detection" feature to find out whether a plant was in the image or not. I think images is an area that could still benefit a lot from machine learning, i.e. facial recognition, image search (extracting meaning from images). thats just some of my ideas. machine learning is awesome I wish you all the best!!
评论 #4093709 未加载
anujkk将近 13 年前
How about creating a recommendation engine for HN articles? It should recommend articles upvoted by users that have similar likes/dislikes as me.<p>I would also like to browse HN articles/comments on basis of topics. For example, right now if I want to find HN posts related to "machine learning" I use HN Search. Can you make it better by using machine learning techniques so that I can get all relevant HN articles on a given topic sortable on basis of upvotes and time?
评论 #4094861 未加载
tzm将近 13 年前
How about analyzing a database of approx 400k car parts requests? Analyze key words, year-make-model, geographic location, etc?<p>I'm currently doing this to generate a chloropleth map for PartsLine.com. <a href="https://skitch.com/tzmartin/eydmm/partsline.com-rfq-to-fips-map" rel="nofollow">https://skitch.com/tzmartin/eydmm/partsline.com-rfq-to-fips-...</a>
ig1将近 13 年前
Why not sign-up to kaggle and do some of their challenges, that way you can also benchmark yourself against others.
评论 #4094833 未加载