★ See also the PDF version of this chapter (better formatting/references) ★
“We make ourselves no promises, but we cherish the hope that the unobstructed pursuit of useless knowledge will prove to have consequences in the future as in the past” … “An institution which sets free successive generations of human souls is amply justified whether or not this graduate or that makes a so-called useful contribution to human knowledge. A poem, a symphony, a painting, a mathematical truth, a new scientific fact, all bear in themselves all the justification that universities, colleges, and institutes of research need or require”, Abraham Flexner, The Usefulness of Useless Knowledge, 1939.
“I suggest that you take the hardest courses that you can, because you learn the most when you challenge yourself… CS 121 I found pretty hard.”, Mark Zuckerberg, 2005.
This is a textbook for an undergraduate introductory course on Theoretical Computer Science. The educational goals of this course are to convey the following:
I hope that following this course, students would be able to recognize computation, with both its power and pitfalls, as it arises in various settings, including seemingly “static” content or “restricted” formalisms such as macros and scripts. They should be able to follow through the logic of proofs about computation, including the pervasive notion of a reduction and understanding the subtle but crucial “self referential” proofs (such as proofs involving programs that use their own code as input). Students should understand the concept that some problems are intractable, and have the ability to recognize the potential for intractability when they are faced with a new problem. While this course only touches on cryptography, students should understand the basic idea of how computational hardness can be utilized for cryptographic purposes. But more than any specific skill, this course aims to introduce students to a new way of thinking of computation as an object in its own right, and illustrate how this new way of thinking leads to far reaching insights and applications.
My aim in writing this text is to try to convey these concepts in the simplest possible way and try to make sure that the formal notation and model help elucidate, rather than obscure, the main ideas. I also tried to take advantage of modern students’ familiarity (or at least interest!) in programming, and hence use (highly simplified) programming languages as the main model of computation, as opposed to automata or Turing machines. That said, this course does not really assume fluency with any particular programming language, but more a familiarity with the general notion of programming. We will use programming metaphors and idioms, occasionally mentioning concrete languages such as Python, C, or Lisp, but students should be able to follow these descriptions even if they are not familiar with these languages.
Proofs in this course, including the existence of a universal Turing Machine, the fact that every finite function can be computed by some circuit, the Cook-Levin theorem, and many others, are often constructive and algorithmic, in the sense that they ultimately involve transforming one program to another. While the code of these transformations (like any code) is not always easy to read, and the ideas behind the proofs can be grasped without seeing it, I do think that having access to the code, and the ability to play around with it and see how it acts on various programs, can make these theorems more concrete for the students. To that end, an accompanying website (which is still work in progress) allows executing programs in the various computational models we define, as well as see constructive proofs of some of the theorems.
This course can be fairly challenging, mainly because it brings together a variety of ideas and techniques in the study of computation. There are quite a few technical hurdles to master, whether it is following the diagonalization argument in proving the Halting Problem is undecidable, combinatorial gadgets in NP-completeness reductions, analyzing probabilistic algorithms, or arguing about the adversary to prove security of cryptographic primitives.
The best way to engage with the material is to read these notes actively. While reading, I encourage you to stop and think about the following:
This book contains some code snippets, but this is by no means a programming course. You don’t need to know how to program to follow this material. The reason we use code is that it is a precise way to describe computation. Particular implementation details are not as important to us, and so we will emphasize code readability at the expense of considerations such as error handling, encapsulation, etc.. that can be extremely important for real-world programming.
This is not an easy course, so why should you spend the effort taking it? A traditional justification is that you might encounter these concepts in your career. Perhaps you will come across a hard problem and realize it is NP complete, or find a need to use what you learned about regular expressions. This might very well be true, but the main benefit of this course is not in teaching you any practical tool or technique, but rather in giving you a different way of thinking: an ability to recognize computation even when it might not be obvious that it occurs, a way to model computational tasks and questions, and to reason about them.
But, regardless of any use you will derive from it, I believe this course is important because it teaches concepts that are both beautiful and fundamental. The role that energy and matter played in the 20th century is played in the 21st by computation and information, not just as tools for our technology and economy, but also as the basic building blocks we use to understand the world. This course will give you a taste of some of the theory behind those, and hopefully spark your curiosity to study more.
This book was initially written for my course at Harvard, but I hope that other lecturers will find it useful as well. To some extent, it is similar in content to “Theory of Computation” or “Great Ideas” courses such as those taught at CMU or MIT. There are however some differences, with the most significant being that I do not start with finite automata as the basic computational model, but rather with Boolean circuits,or equivalently straight-line programs. In fact, after briefly discussing general Boolean circuits and the \(AND\), \(OR\) and \(NOT\) gates, our concrete model for non uniform computation is an extremely simple programming language whose only operation is assigning to one variable the NAND of two others.
Automata are discussed later in the course, after we see Turing machines and undecidability, as an example for a restricted computational model where problems such as halting are effectively solvable. This actually corresponds to the historical ordering: Boolean algebra goes back to Boole’s work in the 1850’s, Turing machines and undecidability were of course discovered in the 1930’s, while finite automata were introduced in the 1943 work of McCulloch and Pitts but only really understood in the seminal 1959 work of Rabin and Scott.
More importantly, the main practical application for restricted models such as regular and context free languages (whether it is for parsing, for analyzing liveness and safety, or even for software defined routing tables) are precisely due to the fact that these are tractable models in which semantic questions can be effectively answered. This practical motivation can be better appreciated after students see the undecidability of semantic properties of general computing models.
Moreover, the Boolean circuit / straightline programs model is extremely simple to both describe and analyze, and some of the main lessons of the theory of computation, including the notions of the duality between code and data, and the idea of universality, can already be seen in this context.
The fact that we started with circuits makes proving the Cook Levin Theorem much easier. In fact, transforming a NAND++ program to an instance of CIRCUIT SAT can be (and is) done in a handful of lines of Python, and combining this with the standard reductions (which are also implemented in Python) allows students to appreciate visually how a question about computation can be mapped into a question about (for example) the existence of an independent set in a graph.
Some more minor differences are the following:
Reducing the time dedicated to automata and context free languages allows instructors to spend more time on topics that I believe that a modern course in the theory of computing needs to touch upon, including randomness and computation, the interactions between proofs and programs (including Gödel’s incompleteness theorem, interactive proof systems, and even a bit on the \(\lambda\)-calculus and the Curry-Howard correspondence), cryptography, and quantum computing.
My intention was to write this text in a level of detail that will enable its use for self-study, and in particular for students to be able to read the text before the corresponding lecture. Toward that end, every chapter starts with a list of learning objectives, ends with a recap, and is peppered with “pause boxes” which encourage students to stop and work out an argument or make sure they understand a definition before continuing further.
roadmapsec contains a “roadmap” for this book, with descriptions of the different chapters, as well as the dependency structure between them. This can help in planning a course based on this book.
This text is constantly evolving, and I am getting input from many people, for which I am deeply grateful. Thanks to Scott Aaronson, Michele Amoretti, Marguerite Basta, Sam Benkelman, Jarosław Błasiok, Emily Chan, Christy Cheng, Michelle Chiang, Daniel Chiu, Chi-Ning Chou, Michael Colavita, Robert Darley Waddilove, Juan Esteller, Leor Fishman, William Fu, Piotr Galuszka, Mark Goldstein, Chan Kang, Nina Katz-Christy, Estefania Lahera, Allison Lee, Ondřej Lengál, Raymond Lin, Emma Ling, Alex Lombardi, Lisa Lu, Aditya Mahadevan, Jacob Meyerson, George Moe, Hamish Nicholson, Sandip Nirmel, Sebastian Oberhoff, Thomas Orton, Pablo Parrilo, Juan Perdomo, Aaron Sachs, Brian Sapozhnikov, Peter Schäfer, Josh Seides, Alaisha Sharma, Noah Singer, Matthew Smedberg, Hikari Sorensen, Alec Sun, Everett Sussman, Marika Swanberg, Garrett Tanzer, Sarah Turnill, Salil Vadhan, Patrick Watts, Ryan Williams, Licheng Xu, Wanqian Yang, Elizabeth Yeoh-Wang, Josh Zelinsky, and Jessica Zhu for helpful feedback.
I will keep adding names here as I get more comments. If you have any comments or suggestions, please do post them on the GitHub repository https://github.com/boazbk/tcs.
Salil Vadhan co-taught with me the first iteration of this course, and gave me a tremendous amount of useful feedback and insights during this process. Michele Amoretti and Marika Swanberg read carefully several chapters of this text and gave extremely helpful detailed comments.
Thanks to David Steurer for writing the scripts (originally written for our joint notes on the sum of squares algorithm) that I am using to produce these notes. David’s scripts are themselves based on several other packages, including pandoc, LaTeX, and the Tufte LaTeX package. I used the [Atom editor] to write these notes, and used the hyrdrogen package, which relies on the Jupyter project, to write code snippets.
Finally, I’d like to thank my family: my wife Ravit, and my children Alma and Goren. Working on this book (and the corresponding course) took much of my time, to the point that Alma wrote in an essay in her fifth grade class that “universities should not pressure professors to work too much”, and all I have to show for it is about 500 pages of ultra boring mathematical text.
Copyright 2018, Boaz Barak.
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.
HTML version is produced using the Distill template, Copyright 2018, The Distill Template Authors.