Machine Learning for Trading CS7646: An Insider’s Review(OMSCS #1)

They say the only way to conquer fear is to face it head-on. So, when I, a wide-eyed newbie, with a non-Computer Science and Engineering background, decided to plunge headfirst into Georgia Tech’s Online Master of Science in Computer Science program, I knew I was embarking on a monumental climb. With a mix of excitement and a hint of trepidation, I envisioned conquering the technical peaks of algorithms and data structures, challenging my mental limits, and embracing a whole new way of thinking.

Now, with my first semester under my belt (and the dust settling from those late-night coding marathons!), I want to share the unvarnished truth about my journey. We’ll delve into the unique tapestry of the OMSCS program, exploring the diverse landscape of courses, the supportive community of fellow climbers, and the ever-present feeling of scaling Mount Everest one byte at a time.

I’ll start by focusing on my first course, Machine Learning for Trading (CS 7646). After reading many reviews on OMS Central, this course, with its promising rating and reputation for being comparatively easier, seemed like the perfect introduction to the program. But I wasn’t content with just dipping my toes in. Fueled by a thirst for knowledge and a touch of overconfidence (or perhaps naivety!), I decided to take on two courses simultaneously: Machine Learning for Trading and Data & Visual Analytics.

Course Overview

The course places a significant emphasis on projects, which account for 71% of the overall assessment. Exams? Yeah, they exist, but chill. Midterms and finals each contribute 12.5%. These exams consist of multiple-choice questions, where each statement is either true or false. Each correct answer is worth 1 point! And there is a total of 5 points for each question (1 point for each statement). With 22 questions in total, you have the opportunity to achieve a maximum score of 110 marks! Scores above 100 are considered extra marks, adding an interesting element to the grading system.

Project Summaries

This course is all about hands-on projects, with different projects counting for different amounts of your final grade.

• Project 1: Martingale (3%). Explore the efficacy of the Martingale betting strategy through probabilistic experiments on an American roulette wheel.
• Project 2: Optimise Something (3%). Evaluate and refine optimization strategies while preparing comprehensive portfolio metrics for enhanced performance.
• Project 3: Evaluate Learners (15%). Implement four Supervised Machine Learning algorithms to assess their effectiveness in diverse applications.
• Project 4: Defeat Learners (5%). Generate datasets to challenge and assess the robustness of Machine Learning algorithms.
• Project 5: Marketsim (8%). Develop a market simulator capable of processing trading orders and tracking portfolio value over time.
• Project 6: Indicator Evaluation (7%). Explore and evaluate various Technical Indicators and formulate a Theoretically Optimal Strategy for effective market analysis.
• Project 7: Q-learning robot (10%). Implement the Q-learning Reinforcement Learning algorithm to develop a responsive and adaptive robot capable of learning and optimizing actions.
• Project 8: Strategy evaluation (20%). Conduct a thorough comparison between a Manual Trading Strategy and an AI-based Strategy Learner, assessing their respective performances in diverse market conditions.

The difficulty of the projects varies, with projects 3 and 8 being significantly more time-consuming. Some projects, such as 2, 4, 5 and 7, do not require the submission of reports, making them relatively less demanding in terms of documentation. It’s worth noting that the weightage may appear uneven, with some seemingly easier projects carrying more weight.

Pro Tips for Project Prowess:

• Start early! Don’t underestimate project demands. Read instructions carefully, they’re your roadmap.
• Report writing matters! Plan and dedicate time to crafting clear and concise reports.
• Rubrics are your friends! Know what gets you points (and what loses them!).

What I Loved:

• Engaging lecture videos by Professor Tucker Balch.
• A wide range of exam preparation resources.
• Responsive Teaching Assistants on the Ed forum.
• Live TA-led discussions to discuss quiz answers and project overviews.
• Flexibility to access all projects from the start.

What Could Be Better:

• Using Gradescope only for code submission, not for grading, meant there was no option to check the marks before the final submission.
• Longer grading periods, which can cause anxiety, especially among newcomers.
• Limited scope for extra credit, the 2% weight for the extra project may not justify its complexity and time required.

Overall Impression

In conclusion, my journey with the OMSCS programme’s Machine Learning for Trading course has been both captivating and enriching. It serves as an excellent entry point for beginners in the field. If you’re well-versed in Python and possess a solid understanding of Machine Learning, pairing it with a more challenging course can be a rewarding endeavour. While achieving an ‘A’ grade is feasible with reasonable effort, delving into supplementary reading materials can offer a more profound comprehension of the subject.

For future OMSCS explorers, this course is a launchpad for your computer science adventure. Pack your curiosity and get ready to level up your skills!