Abstract:
Query complexity, both classical and quantum, are important and well-established
notions of computation. In this thesis, we will investigate query complexity in both
classical and quantum worlds and study their relationship with other closely related
models of computation.
Other than the query model, we will consider a few models of computation, namely
model of computational learning, communications model and local query model for
graphs. We will try to understand the power of these models for various classes of
functions. We will be investigating how various complexity measures (defined as per different
models of computation) relate to each other. In this endeavor we will be working
both upper bound (i.e. designing efficient algorithms in the model of interest) and lower
bound (i.e. proving hardness of computing interesting functions in the model of interest).
As a part of our investigation, we use various mathematical tools like Fourier analysis,
linear algebra, and geometry. Some questions that have motivated the work in this thesis
are:
• How does structural simplicity affect its computational complexity in various models
of computation? Does Fourier analytical simplicity lead to easier computation?
Does geometric simplicity imply efficient communication?
• How does the behavior of quantum computing differ from its classical counterpart?
What is the relation between classical and quantum query complexity of learning
a function? Does the same relationship between a pair of classical complexity
measures also hold in the quantum world?
Understanding the complexity measures that we study and the relationship between
these measures has been an ongoing work for several decades. In this thesis, we push the
boundary of our knowledge a bit further.
The results in this thesis have been divided into three parts.
Part I In this part, we study quantum query complexity from the view of quantum
learning theory. First, we first study the model of exact learning using uniform quantum
examples. Each such quantum example can be generated by making one query to the
function being learned. In this model we are interested in learning the class of Boolean
functions whose Fourier sparsity is bounded, that is, the class of Fourier-sparse Boolean
functions. For this class of functions we give a quantum learning algorithm which improves
upon the tight classical algorithm by Haviv and Regev, 2016. Then, we consider
the model of exact active learning. In this model, a learner is given access to the function
via membership queries. We study the relationship between the number of classical
and quantum membership queries needed to exactly learn a class of Boolean functions,
where improve upon the previous best known relationship by Servedio and Gortler, 2004.
Finally, we study Chang’s lemma, a fundamental result in additive combinatorics. Our
main result here is a refinement of Chang’s lemma for Fourier-sparse Boolean functions.
We also investigate how this lemma is connected to quantum learning theory.
Part II This part is dedicated to the study of the relation between quantum query
and quantum communication complexity. It is well known, in the classical world, that a
query algorithm for a function can be simulated to give a communication protocol of a
closely related communication problem with only a constant overhead. The best known
such simulation theorem in the quantum world, due to Buhrman, Cleve and Wigderson,
1998, has an overhead that is logarithmic in input size of the function. We construct the
first total Boolean function that witnesses this logarithmic gap. This closes a long line
of work. We also give a general recipe of constructing functions that witness separation
between quantum query-to-communication simulation. Finally, we explore the role of
symmetry on this simulation problem.
Part III This part of the thesis is devoted to classical query and communication complexity.
In the first chapter of this part, we explore the role of geometric simplicity, quantified
by bounded Vapnik–Chervonenkis Dimension (VC Dimension) of set systems, in
communication complexity. Our work is motivated by the work of H˚astad and Wigderson,
2007, who considered the Disjointness problem, a canonical problem in communication
complexity, when inputs to the problem are promised to be sets of bounded cardinality.
Indeed, set systems of bounded VC Dimension are a generalization of set systems with
bounded cardinality, with a geometric motivation. Our main result here it to show that
geometric simplicity does not imply efficient communication. In the second chapter of
this part, we consider the problem of estimating the size of the global minimum cut in
the local query model, a fundamental and well-studies model in graph property testing.
We give an algorithm for this problem which almost matches the known lower bound
by Eden and Rosenbaum, 2018. This resolves the query complexity of estimating global
minimum cut in the local query model.