Artificial Intelligence (ARS): Build the Most Powerful AI (Udemy.com)

This has been one of the best courses I have experienced. Make no mistake, the theory and concept behind ARS can be intense and challenging but with persistence, practice and repetition, it will result in a good understanding through the coding of the ARS algorithm itself. It is explained in detail, step by step and in a clear and informative way specifically throughout the entire coding phase as to leave no ambiguity. The best part of it is when you can see your persistence pay off in a working program and the animations produced while training your AI and watching how it improves itself and how its also possible to play with the parameters to get some very interesting results.

Nice, Nice experience. very good course. Can I adapte this algo for decision-making? for example, 1. energy efficient consumption of specific equipment in a building, for instance deciding for jettison? or 2. for deases diagnostic? by collecting data from sensors network made with Arduino and Raspberry? also I could not instal pybullet, I try all the versions but Iam very satisfied . building an AI from scientific corpus is very exciting, usefullness

I like Kirill and Hadelin as teachers. I think their separation of intuition and practical tutorials is a very good technique and executed well. I have one complaint for this course though. I believe it needs a few more videos at the end... In comparison to the Super Data Science Team's other courses, this one has a narrower topic (one AI in particular) and therefore should go into more depth. I think it would have been appropriate to discuss the interface of the AI with the environment more thoroughly so students would know how to apply it to more situations. I am trying my own little challenge on this AI and I am discovering architectural changes I need to make to the AI code created in this course for it to handle my environment. Secondly, we didn't even get to see the half cheetah video of the reward policy that achieved > 900. This just left me with an incomplete feeling at the end. None-the-less, I'm glad I purchased.

I really loved this course, it provide a step by step guide of everything we need to know and walks us by the entire algorithm described in the research paper. I like the explanations and that they take their time to describe everything correctly. I'm grateful that they found a way to merge pybullet with anaconda so we don't lose time in trying to get it working.

good course on how to read a research paper, and convert the math from the paper to Python, creating a Percepton, and the methods around a newly published shallow learning technique. Additionally learn dozens of python, coding of basic statistical routines, (running mean, variance, standard deviation, and inserting loops into dictionaries for sorting! ). ... Both Kirill and Hadelin bring a delightful approach to the AI that they find exciting. Hadelin is efficient in presenting how to download Anaconda, and setup the first file, and introducing the spyder ide. Furthermore, the tutorials progress rapidly, richly annotated by Hadelin's providing you an over his shoulder view, as Hadelin explains each line of code. This course is worth multiple times the offering price. Within several hours you'll be up and running the very latest in AI .

Real would examples of ARS implementation would be valuable. It seems like only one half of the instuction needed to make ARS do something useful is provided. It is left to the student to figure out how to structure and implement your Ai project to work with ARS since no examples are provided on the source data side. Otherwise, I found the intutive explanation and practical coding split course setup easy to follow and informative.

The subject matter was great--very practical and very cutting edge and valuable. The way the course was laid out was great as well. It is, of course, for someone a bit advanced in AI. What I might recommend is to talk a little more naturally as you would talk to a friend rather than trying to pause as though you were attempting to prompt the student to give an answer. For example, "Instead of using a?....(long pause).... for loop in this situation, since you are involving?...(long pause)..... several variables, you may want to try a?.....(long pause).....a while loop." This technology is really new, so it really is not fair to expect the instructor to have exercises laid out and tested. But it might be something you could add to the course to make it a bit more complete or educational. Regardless, I am *very* glad I took this course and I expect to benefit from it greatly. Thank you so much for making it available.

This was my first crack at AI. I felt this course really helped to understand the mathematical concepts AI at a high level. The initial intuitive lectures offered a am array of business cases where I can implement the AI built. The technical portion was easy to follow and delivers the promised asset. However, there was no intuitive follow-up after completion outlining basic steps to adapt what was built. There could have been a better closing and bringing of the two aspects of the course (intuitive and technical) together. Preferably by including an example of using the AI in another environment specific to one mentioned in the intuitive portion.

Paradoxically what I got from the course is not the ability to build "the most powerful AI" but rather: - setting up OpenAI Gym with PyBullet environment (which I have never done before) and building my first reinforcement learning model - sharpening my Python skills a bit - a bit of "can do" attitude when it comes to approaching research papers. The ASG learning method proposed is fast, but its applicability is very limited - it is in the end just a linear model, with naturally limited applicability. Can't wait to experiment with extending it. Having said that - I enjoyed this course enormously and seeing your first "Half Cheetah" learn to walk is extremely satisfying experience. Thanks and Enjoy AI :-)

An excellent course - Definiely recommend it to people with some prior exposure to Machine Learning and Reinforcement Learning. Really appreciate the fact that SuperDataScience Team came up with a course relating to such a recent Research Paper. Looking forward to more courses from Udemy that keeps the ML Community in touch with the latest advancements in ML/RL. Having said that- I wish the course had some coverage on the following topics: 1. Is ARS a good fit for RL problems that deal with discrete actions ? OR is it only suited for continuous control problems ? 2. What exactly do the elements of the state space and action space of the half-cheetah problem represent ? 3. What if the model output values for the action space that are NOT in the correct / legal range of values ? Maybe some of these were covered OR obvious - but I am still trying to figure out the answers to these.

Obvious that instructor knows what he is talking about and his enthusiasm is contagious. Really looking forward to learning about ARS

interesting.