This book and control theory solve different problems. I'm just commenting because I sometimes see modern ML folks trying to apply super duper general techniques to problems that are easily solved specifically with simple undergrad-level techniques from the specific fields (including this material to control problems). Also, I just think control theory is really cool and perhaps under-discussed in "cool math/eng stuff" circles.
Also, unfortunately, classical control theory is primarily good for linear time-invariant dynamics in the frequency space of the Laplace transform. If you can't locally linearize your model and/or need to learn a model, classical control approaches are underdeveloped, and everyone has switched to optimal control and RL.
A brief skim through of the book shows many references to Control Theory texts.
It's a rather technical book, focusing on issues like advanced sensor data fusion, control theory, and optimization. Not that there is anything wrong with that, the title is just slightly misleading.
Wendell H. Fleming and Raymond W. Rishel,
Deterministic and Stochastic Optimal
Control, ISBN 0-387-90155-8,
Springer-Verlag, Berlin, 1979.
There is more on optimization, especially
from the Hahn-Banach theorem, and also
Kalman filtering, in
David G. Luenberger, Optimization by
Vector Space Methods, John Wiley and
Sons, Inc., New York, 1969.
See his blog post "What we've Learned to control"  and the survey paper "A Tour of Reinforcement Learning: The View from Continuous Control"  may be of interest.
it's not a single tool but more of a framework. it goes beyond frequency based representations and can also model non-linear control problems.
the kalman filter is an example of the use of this framework where it's combined with statistics.
the problem in control theory though is that once you want to go beyond linear control things get very difficult, most of the literature seems to be about finding clever ways of approximating your problem into something linear.
PS for others, it's on libgen...
This associate professor of Decision Theory received two offers of tenure track positions at reputable universities, and got some friends and colleagues together to discuss which one to take. One of them suggested he use the tools of his discipline, Decision Theory, to help him make a decision. To which the professor replies: "Now, come on guys, this is serious..."
Here is the direct link if you don't care about my stuff (it has been featured on HN already)
> One of my most endearing memories of the great psychol- ogist of decision making under uncertainty, Amos Tversky, recalls his way of ordering in restaurants: “Barbara? What do I want?”
Julia is an awesome language and the ecosystem around it is getting better quickly.
- Get very familiar with the table of contents first. Then note and filter out the sections you are less interested in or don't appear to provide the biggest bang for your buck.
- Annotate, annotate, annotate. Active reading is better than passive reading. I use a 2-in-1 chromebook and a passive pen to highlight in the adobe pdf reader.
- Write detailed notes in addition to annotating (i.e. in a separate notebook or markdown file). Cull/convert your notes into Anki/Memrise digital flashcards. Drill on your phone when you are lying in bed.
1. Ten pages a day
2. Do all the exercises, and redo all the ones you get wrong.
3. Add anki cards for all the key terms, ideas, formulas, etc. The best textbooks include a summary the end of every chapter that you can double check your notes / anki against. Prefer cloze markup to question / answer cards -- more cards per "fact" but longer retention.
4. Double check your cards against a source like Wikipedia for completeness / accuracy, since sometimes there's an author bias to content with.
Thought that was really interesting.
But gotta love goggins, dude likes taking the hardest path to success.
Quite a remarkable level of determination.
But, the focus is not on internalising the material, the focus is on producing something that makes sense in a different language.
I guess part of the problem is that the source material I am working from is from 1596, in blackletter, and pretty archaic "source language", so the actual process is "transcribe original", "translate transcription".
so I would say, try to find something you want to do that requires the information in the textbook. or else just don't bother reading the textbook and go do something else you actually want to do.
Much more up-to-date than sutton and barto, but the authors are rather less well known.
It was one of my favorite courses, and it looks like this textbook is now used for that course:
however I wish it also described the dual of this linear program. this problem involves optimizing over state-action frequencies which is equivalent to optimizing over policies.
so value functions and policies are dual to each other. that's pretty neat! not sure why modern RL texts don't talk about it at all.
Authors, if you're reading this, thanks a thousand for making this available to us!
I sound so annoyed because I recently downloaded 'hist.pdf', 'fxtbook.pdf' and 'V090212S.pdf' Is it just to get a memorable file URL? If so, humanity invented simlinks long ago.
Making the main file named something more verbose would impact you whenever you try to open it, but in an ideal world, you'd rename it once it gets to the point of it not needing many changes.
I'm trying to be considerate and provide a uniquely identifiable file name whenever I make some available to the public. In my opinion, books require the full name and revision, and, as I've already mentioned, you can have a link to it (or a copy) with a short name if you'd like to have an option of a short URL for verbal link transmission.