To accelerate research on adversarial examples and robustness of machine learning classifiers, Google Brain organized a NIPS 2017 competition that encouraged researchers to develop new methods to generate adversarial examples as well as to develop new ways to defend against them. In this chapter, we describe the

No abstract

We explore the intricate interdependent relationship among counting problems, considered from three frameworks for such problems: Holant Problems, counting CSP and weighted H-colorings. We consider these problems for general complex valued functions that take boolean inputs. We show that results from one framework can be used to derive results in another, and this happens in both directions. Holographic reductions discover an underlying unity, which is only revealed when these counting problems are investigated in the complex domain C. We prove three complexity dichotomy theorems, leading to a general theorem for Holant c problems. This is the natural class of Holant problems where one can assign constants 0 or 1. More specifically, given any signature grid on G = (V, E) over a set F of symmetric functions, we completely classify the complexity to be in P or #P-hard, according to F , ofwhere fv ∈ F ∪ {0, 1} (0, 1 are the unary constant 0, 1 functions). Not only is holographic reduction the main tool, but also the final dichotomy can be only naturally stated in the language of holographic transformations. The proof goes through another dichotomy theorem on boolean complex weighted #CSP.

Holant problems capture a class of Sum-of-Product computations such as counting matchings. It is inspired by holo-graphic algorithms and is equivalent to tensor networks, with counting CSP being a special case. A complexity classification for Holant problems is more difficult to prove, not only because it logically implies a classification for counting CSP, but also due to the deeper reason that there exist more intricate polynomial time tractable problems in the broader framework. We discover a new family of constraint functions L which define polynomial time computable counting problems. These do not appear in counting CSP, and no newly discovered tractable constraints can be symmetric. It has a delicate support structure related to error-correcting codes. Local holographic transformations is fundamental in its tractability. We prove a complexity dichotomy theorem for all Holant problems defined by any real valued constraint function set on Boolean variables and contains two 0-1 pinning functions. Previously, dichotomy for the same framework was only known for symmetric constraint functions. The set L supplies the last piece of tractability. We also prove a dichotomy for a variant of counting CSP as a technical component toward this Holant dichotomy.