There are some lectures on RL and some on Trading. But there aren't enough materials on the application of RL to Trading. It just talks about some high level concepts on how it could be used. We could get this from any basic article on RL and Trading. Even the last exercise is not RL on Trading. It's just a machine learning exercise to predict S&P500's direction. Basically there is zero example and exercise on RL for Trading Strategies, which is the main topic.

This course could have been better if a full trading strategy project using Reinforcement Learning is provided, but it just provides some theories and sample Reinforcement Learning code instead.

The course could be improved if the lab included stock trading related works for both RL and LSTM. I had already learned stock trading with RL and LSTM before I took this class.

Reinforcement learning tasks are not related to financial domain. Financial topics are superficial. Course for absolute newbies in RL and FinTech

Good for begginer, but this course was heavy binded with GCP which might not practical in some country.

The labs need an update, otherwise interesting mix of content. It was useful, but can be better

no actual model on stock prediction using RL

Good and bad:

Good: Jack Farmer's explanations about the financial parts of the course were great. Very well explained and useful.

Bad: The labs and technical part of the course were horrible... I would say half of the labs did not work properly, I couldn't finish them or sometimes even go beyond the 3rd cell of code because there were many errors that I did not know how to fix (we are not supposed to do that). The final project is supposed to let us create our own model but we haven't been taught enough to use the different tools since the labs do not teach you enough.

All in all, I only give the 2 stars as an average because Jack's part was on point, you really learn useful insights, 5/5, but the labs were beyond bad (0/5).

There is no real application of RL in trading in this course. They just first skim quickly to the basics of RL, quite superficially, then they explain the basics of portfolio management. These two rails go parallel and never touch each other. Moreover, the part covering RL, MDP, TD and Q Learning is illustrated too fast to understand any subtle points, with too many details (equations quickly explained, code fragments gone through in a minute or too) put together roughly to be a qualitative introduction.

I appreciated the general theoretical components, exposed in a clear and concise way. However, I found the total lack of in-depth AI applications on real cases, which are essential for using the notions in practice. From a concrete point of view, therefore, the course was of no use to me.

A few interesting nuggets buried in a mess of cobbled together material, dodgy slide decks with poorly formatted code snippets, all combined with the annoying "QwikLabs" that takes about 3 minutes to start for every single assessment. This could be so much better.

Giving it two stars because the material was interesting. HOWEVER, it does not cover how to use reinforcement learning for trading strategies. It discusses reinforcement learning, and trading, but fails to connect the dots. This course feel incomplete.

Way below expectations!

A good chunk of the practical labs are extremely outdated - and are so faulty that the parts involving the Google Cloud Platform fail just about in every lab. You might as well just download the freely available jupyter workbooks and follow along visually, because fixing them to run them as GCloud changes their interface is a thankless task that should have be done by the people who charge for this course.

The specialization needs a major revamp. It almost useless as it stands. The topics covered are important but the labs are horrific. The Aquan toolbox used is not maintained by anyone anymore and had I had to fix it myself. The RL topics and the LSTM are covered in haste, like going through the motions.

80% of the course is generic DL without any connection to RL in trading, NLP / image processing and more unrelated topics.

It's just a group of unrelated lectures aggregated in to a "course". its the first coursera course that I took and the last one for sure.

No useful information regarding the RL for trading was shared. It does not provide any information regarding the feature engineering, understaing the required input features, and so on. The RL parst were also at a very beginner level

The labs are junk. Half of them are broken, some of them are missing. Those that are working, are not good enough. Week 3, AutoML part doesn't mention trading even once. The time you will spend on this course doesn't worth it.

Too general to pursue any meaningful work with RL for trading. The class is trying to cover too much material from too many different angles to be useful.

Cannot understand the girl in https://www.coursera.org/learn/trading-strategies-reinforcement-learning. Had to leave on subtitles. Very distracting.

Highly limited information with extremely steep learning curve.

Bozo education at its best!

Labs Malfunction

你对得起我们吗？日内瓦退钱

It's an exceptionally difficult task to predict financial time series, and even harder to design an automated trading methodology that can take into account those forecasts while monitoring the trading environment (trading costs, other traders, sentiment). This final course is an ambitious attempt to expose learners to the most advanced concepts in the field.

To be able to comprehend the Reinforcement Learning materials appears to require expertise in deep learning far beyond sequential models, and also appears to need the volume and integrity of data only available to high-frequency trading firms. Thumbs up to the specialization curators for providing a non-trivial introduction.

The module that rescued the course (and lifted my rating to 5 stars) was the AutoML demonstration. I was reassured to see that Gradient Boosted Trees were chosen as the appropriate methodology, as this is what I have casually observed as being the most effective methodology in use today for end-of-day data problems. Looks like an amazing product, if you have the money!