Next time when you use Amazon S3 service, don’t take it for granted, because that’s very successful engineering work, compared with who? Ceph. Now I’m talking about Ceph S3 (CephFS is another topic that will not be covered here).
Automata itself is a very interesting thesis and Turing Machines which Automata leads to are breathtakingly beautiful.
This is my first time to learn Quantum Mechanics, Quantum Mechanics is a very cool discipline to me.
A handy text editor should have these functions built in or have plugins making up the missing piece.
Sometimes you are sleeping after midnight, but your neighbors are excited talking loudly, or watching a noisy show, or even continuing their party. What can you do?
Competitive Programming can sharpen one’s mind, sometimes helps with one’s interview. This post is a note of Bjarki Ágúst Guðmundsson’s Competitive Programming course.
GitHub Pages is a great place to host static content for personal use, especially with the help of Jekyll, and it’s accessible in China. GitHub Pages also supports custom domain, and HTTPS protocol, but not simultaneously, if you configured your custom domain for your GitHub Pages, you can only use HTTP. Maybe one day GitHub will let you HTTPS your own domain on it, but for now you have to get around it.
Books about China, China’s economy, politics, environment.
I was afraid of designing, I didn’t know what to think when asked to design, my design vocabulary was too small. After this course I think I’m still naive in designing, but at least I’d be a less scared designer and really I’ve grabbed some small tricks.
TodoPomo is a Visual Studio Code extension that combines To Do Task and Pomodoro. It helps you focus on 25-mins work period and track your everyday achievements.
Finally a non-science course, purely a technical one, I can’t wait to make my hands dirty on web development. I think every developer whether he/she is front end or back end knows web development more or less, but to make your site/single page application professional and pretty, you have to go through some courses systematically.
Physics is my favorite discipline, but it’s been so many years since I graduated from high school when it was the last period of the time I acquired new physics knowledge. I have various excuses not to pick up physics learning and gain new physics knowledge these years, I’m too busy, too tired, have no time, have other immediately required skills to learn, etc. Physics didn’t find its way to my priority. But it’s not he case this time, my laptop was sent back to Apple Store to get repaired, all I have in hand now are a tablet, a phone, a paper notebook and a pencil. I can’t code, can’t blog, I can only read something and write down something by hand, physics gets its chance today.
There are two version of this course in CMU, the other one is “Probability & Statistics”, I prefer this version because the “probability” part in this version acts as a “bridge” to the inference section thus making easier to understand and grab the big picture.
This is not an exhaustive reference for R language, I just wrote this according to my recalling of DataCamp courses and Data Science Specialization by Johns Hopkins University on Coursera, but this may be suitable for most data analysts and data scientists.
Recently I read a memoir written by Li Zhisui, one of the physicians to Mao Zedong. Li Zhisui worked for Mao from 1954 to 1976 (when Mao died of ALS), totally for 22 years. He had a close view on Mao’s private life and told us many details about Mao’s political motives to start the Great Leap Forward and The Cultural Revolution, the two great disasters Mao brought to China.
One day I haven’t spent totally before computer, and I get red-necked now.
Some tools and configuration examples can be found in this GitHub repository.
Installing OS is a bad job, rescuing one is worse, recovering data is the worst. But most time when it comes, we have to face it.
This is a cheat sheet for Recurrent Neural Network learner, covers from math essentials of neural network to several different kinds of RNN. If you go through the whole content and understand thoroughly, you will at least be able to make your own neural network architecture, derive forward/backward propagation equations, which are fundamental to building your own working neural network.
FancyWord is a Sublime Text 3 dictionary plugin that improves your word choice in English writing.
FancyWord 是一个辅助用户在英文写作中更好地选词的 SublimeText 插件。
Update on 7/16: This course is damn hard! And poor organized somewhere, for example, in week 13 one video is missing and the last two questions in the quiz are wrong themselves. But all in all it’s still a course in depth.
This is a note for Course: Neural Networks for Machine Learning University of Toronto
I found Prof. Geoffrey Hinton’s British English was a little hard for me to understand, but he definitely has the insight of neural network, the content is really of high quality and helped me a lot to understand neural network thoroughly.
Today I was trying a RNN model using Keras and TensorFlow, and some warnings just came out:
On Mac, you have many ways to install Python and its packages: pip, conda, brew, port, offical source or binary, etc. But none of them is perfect, sometimes you find you can only install a package through like conda or brew, but after installation, you start your port Python, but can’t find that package at all.
This is a note of a Deep Learning course I just got a certificate from: Deep Learning with TensorFlow by BigDataUniversity.
This course can only give you a simple taste of how to use TensorFlow and part of essential concepts like tensor, variable, placeholder, operation, layer, etc. It can’t teach you the underlying mathematical principles or how to design a reasonably stacked neural network. Besides this course I still encourage you attend Machine Learning, Neural Network courses to deepen your understanding and enable yourself wisely organize your network architecture.
When I use Sublime Text to edit Markdown files, I appreciate the Syntax Highlight provided by Markdown Light plugin, but just like built-in and other Markdown plugins, they can’t provide auto completion for embedded LaTex statements in Markdown files. But it’s very common nowadays to mix various branches of Markdown with LaTex-like math equations, it will ease my pain if I can have LaTex auto completion when editing Markdown files.
This course is one of the most famous courses on Coursera. Now I go two weeks ahead of the deadline and reach Week 5, I plan to finish it in the flowing few days.
When you’re tweaking Python’s package manager PIP, make sure you know these.
I use R with Python a lot, Octave is the chosen language in Coursera course: Machine Learning by Stanford University.
So this article will only cover necessary concept to finish this Machine Learning course.
Easy way to tell which Init System you are using? And how to manipulate services in different Init System.
Python enable us to perform advanced operation in very expressive way, meanwhile covers many users’ eyes from underlying implement details. If the performance of your application plays a critical role, please always keep in mind the time complexity of these common operations.
I’ve finished a Data Science course on Coursera recently, Data Manipulation at Scale: Systems and Algorithms, which is taught by Bill Howe.
Sublime Text rocks, but can you run Python projects interactively while using its Build System?
Recently I find Anaconda plugin for Sublime Text 3 has some performance issue. If your buffer is wider than like 20,000 characters, or higher than 5,000 lines, Anaconda will fail, the Autocomplete, Docstring and Goto_definition function will have several seconds’ lag or will take forever before they can show up, sometimes Anaconda’s jsonserver.py process even takes 100% CPU usage on one core and fails quiting when Sublime Text 3 main process is terminated.
Before I moved to California, all meal-replacements I could buy are too expensive, so I made my own at home, and here was my recipe if you can read Chinese.
Now I have more options: Soylent, Joylent, 100%FOOD, etc. And most times, they are cheaper than Chinese similar products. Soylent and 100%FOOD are based in USA and Joylent is in England.
I have an idea recently: predict Bitcoin’s future price, and use that predication to assist my trading decision, and make profit from it. To tell if this is profitable and how profitable in academic terms is hard for me, but do it is not a big deal for me. Let’s just build one Bitcoin Trader bot.
I am so lazy, I hate spending time on eating, but I have to spend 0.5-1 hours to have meals every day. What a shame and a waste of time for lazy people like me, I can name 1,000 more important things than having regular food. Maybe we can DIY our meals which will take us less than 2 minutes and meanwhile meet our nutritional requirements, it also sounds good if there already exist such products.
Fuxi is a symbolic reasoning engine writen in Python.
subscribe via ATOM