I have just started a course on artificial intelligence. The course opens with an assignment where the student must implement the minimax algorithm and variations on the algorithm in order to play a reversi-like game. The game is called isolation and is very similiar to tron. Players make a move, rendering the previously occupied square unusuable for the duration of the game. The objective of the game is to prevent the opponent from making anymore moves.
I just took a course on software analysis. The course was created by Prof. Mayur Naik and the course website is here. Course notes are available here. Below I present some facts I learned. The syllabus lists a bunch of tools used for software analysis. I am going to discuss randoop type systems dynamic symbolic execution Randoop Here is some background reading material on the randoop tool. Some employees at Microsoft Research worked on a technique for test generation called “feedback-directed random test generation”.
Stock market data is usually prices for a particular stock. For each epoch of time, the max, min, volume of trade, and final price for the day is listed. An adjusted price may be listed at extra charge (accounting for stock splits). More information can be derived from these prices through the use of technical indicators (TI). TIs are measures to describe the trading activity of a security. The following are examples of properties TIs measure:
I just took a class on machine learning for trading and now I'm itching to set up my own algorithms and make bank. The first problem I need to solve is getting data. My choices are: Quandl $30 a month for US equities data Alpaca.markets open brokerage account, get free data commision free trading for stock and etf Interactive Brokers open brokerage account, get free data commision free trading for stock and etf (see IBKR lite) AlphaVantage free data on stocks, forex, TIs rate limit 5 api calls a minute, max 500 calls a day I'm going to go with alpaca.
I host this blog using Github Pages. Some resources such as CSS and JS are hosted external and others with the same origin. I encountered some CORS errors with CSS files hosted on Github Pages. There were no “Access-Control-Allow-Origin: *” HTTP headers being passed. The Hugo template I am using, hugo-flex, adds crossorigin attributes when loading the base css stylesheet. Github uses Varnish for serving up pages. Removing the attributes let the CSS files be served without CORS error.
Looks like I was wrong about static site generators. There's a bit of learning curve, but after figuring out the templating language and how to get latex working, I realize the value of such tools. I'll continue to use tiddlywiki, but not to host a blog. A 3 mb html file seems irresponsible to host, considering most of the world does not enjoy low cost bandwidth yet. Now, unless I invent more problems for myself, I should get on with whatever I was procrastinating from.
I just spent an hour trying to figure out a good solution for MathML rendering in Hugo. I could not figure it out, it will likely take a couple more days to click. In the mean time, I need to figure out how to change the formatting in this TiddlyWiki to match this theme. The TiddlyWIki theme is not very easy to read. The text needs to be centered. If all else fails, I'll just have to write my static web pages manually.
I have been thinking about learning the Go and Rust programming languages, but I am not too sure what their purposes are. I've been meaning to build projects for my portfolio. I have considered using Python or Java. I would like to build a server application that can scale very well. For that reason, I am considering learning Go. But first, I should survey what other people think of Go and Rust.
This has been an exciting summer. Machine learning has been quite the buzzword in the past few years. Now that I have learned more about the methods, I know for a fact that they hype is a bunch of crap. Garbage in garbage out. If an organization has bad data, machine learning will garner terrible insights from that data. some ML techniques: LinearRegression DecsionTree RandomTree BagLearner k-Means QLearning some domain knowledge on markets
Hoo-wee. Up to assignment 4 for the Machine Learning for Trading course. A quick summary of the previous three. Martingale betting strategy formulation Spin the wheel, win at roulette. Evaluate betting strategy Basic probability (n out of m slots) Optimize a portfolio to maximize sharpe return Useful methods for evaluating portfolios of securities Implement and assess decision trees and boostrap aggregating methods machine learning basics decision trees can double as classifier and regressor