
We solve the optimal control problem of a onedimensional reflected
stochastic differential equation, whose coefficients can be path dependent. The
value function of this problem is characterized by a backward stochastic
partial differential equation (BSPDE) with Neumann boundary conditions. We
prove the existence and uniqueness of sufficiently regular solution for this
BSPDE, which is then used to construct the optimal feedback control. In fact we
prove a more general result: The existence and uniqueness of strong solution
for the Neumann problem for general nonlinear BSPDEs, which might be of
interest even out of the current context.

In [Azimzadeh, P., and P. A. Forsyth. "Weakly chained matrices, policy
iteration, and impulse control." SIAM J. Num. Anal. 54.3 (2016): 13411364], we
outlined the theory and implementation of computational methods for implicit
schemes for HamiltonJacobiBellman quasivariational inequalities (HJBQVIs).
No convergence proofs were given therein. This work closes the gap by giving
rigorous proofs of convergence. We do so by introducing the notion of nonlocal
consistency and appealing to a BarlesSouganidis type analysis. Our results
rely only on a wellknown comparison principle and are independent of the
specific form of the intervention operator.

We use the Markov chain approximation method to construct approximations for
the solution of the mean field game (MFG) with reflecting barriers studied in
Bayraktar, Budhiraja, and Cohen (2017). The MFG is formulated in terms of a
controlled reflected diffusion with a cost function that depends on the
reflection terms in addition to the standard variables: state, control, and the
mean field term. This MFG arises from the asymptotic analysis of an $N$player
game for single server queues with strategic servers. By showing that our
scheme is an almost contraction, we establish the convergence of this numerical
scheme over a small time interval.

This paper studies the utility maximization on the terminal wealth with
random endowments and proportional transaction costs. To deal with unbounded
random payoffs from some illiquid claims, we propose to work with the
acceptable portfolios defined via the consistent price system (CPS) such that
the liquidation value processes stay above some stochastic thresholds. In the
market consisting of one riskless bond and one risky asset, we obtain a type of
superhedging result. Based on this characterization of the primal space, the
existence and uniqueness of the optimal solution for the utility maximization
problem are established using the duality approach. As an important application
of the duality theorem, we provide some sufficient conditions for the existence
of a shadow price process with random endowments in a generalized form as well
as in the usual sense using acceptable portfolios.

Oftcited causes of miniflash crashes include human errors, endogenous
feedback loops, the nature of modern liquidity provision, fundamental value
shocks, and market fragmentation. We develop a mathematical model which
captures aspects of the first three explanations. Empirical features of recent
miniflash crashes are present in our framework. For example, there are periods
when no such events will occur. If they do, even just before their onset,
market participants may not know with certainty that a disruption will unfold.
Our miniflash crashes can materialize in both low and high trading volume
environments and may be accompanied by a partial synchronization in order
submission.
Instead of adopting a classicallyinspired equilibrium approach, we borrow
ideas from the optimal execution literature. Each of our agents begins with
beliefs about how his own trades impact prices and how prices would move in his
absence. They, along with other market participants, then submit orders which
are executed at a common venue. Naturally, this leads us to explicitly
distinguish between how prices actually evolve and our agents' opinions. In
particular, every agent's beliefs will be expressly incorrect.

We propose notions of minimax and viscosity solutions for a class of fully
nonlinear pathdependent PDEs with nonlinear, monotone, and coercive operators
on Hilbert space. Our main result is wellposedness (existence, uniqueness, and
stability) for minimax solutions. A particular novelty is a suitable
combination of minimax and viscosity solution techniques in the proof of the
comparison principle. One of the main difficulties, the lack of compactness in
infinitedimensional Hilbert spaces, is circumvented by working with suitable
compact subsets of our path space. As an application, our theory makes it
possible to employ the dynamic programming approach to study optimal control
problems for a fairly general class of (delay) evolution equations in the
variational framework. Furthermore, differential games associated to such
evolution equations can be investigated following the KrasovskiiSubbotin
approach similarly as in finite dimensions.

We consider an nplayer symmetric stochastic game with weak interaction
between the players. Time is continuous and the horizon and the number of
states are finite. We show that the value function of each of the players can
be approximated by the solution of a partial differential equation called the
master equation. Moreover, we analyze the fluctuations of the empirical measure
of the states of the players in the game and show that it is governed by a
solution to a stochastic differential equation. Finally, we prove the
regularity of the master equation, which is required for the above results.

We introduce high order Bellman equations, extending classical Bellman
equations to the tensor setting. We show that a sufficient condition for the
existence and uniqueness of a positive solution is that the tensors appearing
in the equation are strong Mtensors. We give a policy iteration algorithm to
compute the positive solution. We also extend the notion of weakly chained
diagonal dominance to tensors. We prove that a weakly diagonally dominant
Ztensor with nonnegative diagonals is a strong Mtensor if and only if it is
weakly chained diagonally dominant. This is analogous to a corresponding result
in the matrix setting and tightens a result from [Z. Liping, L. Qi, and G.
Zhou. "Mtensors and some applications." SIAM Journal on Matrix Analysis and
Applications (2014)]. We apply our results to obtain a provably convergent
numerical scheme for an optimal control problem using an "optimize then
discretize" approach which outperforms (in both computation time and accuracy)
a classical "discretize then optimize" approach. To the best of our knowledge,
a link between Mtensors and optimal control has not been previously
established.

Let $\Omega$ to be one of $\X^{N+1},C[0,1],D[0,1]$: a product of Polish
spaces, space of continuous functions from $[0,1]$ to a subset of
$\mathbb{R}^d$, and space of RCLL (rightcontinuous with left limits) functions
from $[0,1]$ to $\mathbb{R}^d$ respectively. We consider the existence of a
probability measure $P$ on $\Omega$ such that $P$ has the given marginals
$\alpha$ and $\beta$ and satisfies some other convex transport constraints,
which is given by $\Gamma$. The main application we have in mind is the
martingale optimal transport problem with when the martingales are assumed to
have bounded volatility/quadratic variation. We show that such probability
measure exists if and only if the $\alpha$ average of socalled $G$expectation
of bounded uniformly continuous and bounded functions with respect to the
measures in $\Gamma$ is less than their $\beta$ average.
As a byproduct, we get a necessary and sufficient condition for the Skorokhod
embedding for bounded stopping times.

Inspired by Strotz's consistent planning strategy, we formulate the infinite
horizon meanvariance stopping problem as a subgame perfect Nash equilibrium in
order to determine time consistent strategies with no regret. Equilibria among
stopping times or randomized stopping times may not exist. This motivates us to
consider the notion of liquidation strategies, which lets the stopping right to
be divisible. We then argue that the meanstandard deviation variant of this
problem makes more sense for this type of strategies in terms of time
consistency. It turns out that an equilibrium liquidation strategy always
exists. We then analyze whether optimal equilibrium liquidation strategies
exist and whether they are unique and observe that neither may hold.

Since most of the traded options on individual stocks is of American type it
is of interest to generalize the results obtained in semistatic trading to the
case when one is allowed to statically trade American options. However, this
problem has proved to be elusive so far because of the asymmetric nature of the
positions of holding versus shorting such options. Here we provide a unified
framework and generalize the fundamental theorem of asset pricing (FTAP) and
hedging dualities in arXiv:1502.06681 (to appear in Annals of Applied
Probability) to the case where the investor can also short American options.
Following arXiv:1502.06681, we assume that the longed American options are
divisible. As for the shorted American options, we show that the divisibility
plays no role regarding arbitrage property and hedging prices. Then using the
method of enlarging probability spaces proposed in arXiv:1604.05517, we convert
the shorted American options to European options, and establish the FTAP and
sub and superhedging dualities in the enlarged space both with and without
model uncertainty.

This paper studies the solvability of a class of Dirichlet problem associated
with nonlinear integrodifferential operator. The main ingredient is the
probabilistic construction of continuous supersolution via the identification
of the continuity set of the exit time operators under Skorohod topology.

With model uncertainty characterized by a convex, possibly nondominated set
of probability measures, the agent minimizes the cost of hedging a path
dependent contingent claim with given expected success ratio, in a
discretetime, semistatic market of stocks and options. Based on duality
results which link quantile hedging to a randomized composite hypothesis test,
an arbitragefree discretization of the market is proposed as an approximation.
The discretized market has a dominating measure, which guarantees the existence
of the optimal hedging strategy and helps numerical calculation of the quantile
hedging price. As the discretization becomes finer, the approximate quantile
hedging price converges and the hedging strategy is asymptotically optimal in
the original market.

We consider a large queueing system that consists of many strategic servers
that are weakly interacting. Each server processes jobs from its unique
critically loaded buffer and controls the rate of arrivals and departures
associated with its queue to minimize its expected cost. The rates and the cost
functions in addition to depending on the control action, can depend, in a
symmetric fashion, on the size of the individual queue and the empirical
measure of the states of all queues in the system. In order to determine an
approximate Nash equilibrium for this finite player game we construct a
LasryLions type meanfield game (MFG) for certain reflected diffusions that
governs the limiting behavior. Under conditions, we establish the convergence
of the Nashequilibrium value for the finite size queuing system to the value
of the MFG.

In the multisensor sequential change detection problem, a disruption occurs
in an environment monitored by multiple sensors. This disruption induces a
change in the observations of an unknown subset of sensors. In the Byzantine
version of this problem, which is the focus of this work, it is further assumed
that the postulated changepoint model may be misspecified for an unknown
subset of sensors. The problem then is to detect the change quickly and
reliably, for any possible subset of affected sensors, even if the misspecified
sensors are controlled by an adversary. Given a userspecified upper bound on
the number of compromised sensors, we propose and study three families of
sequential changedetection rules for this problem. These are designed and
evaluated under a generalization of Lorden's criterion, where conditional
expected detection delay and expected time to false alarm are both computed in
the worstcase scenario for the compromised sensors. The firstorder asymptotic
performance of these procedures is characterized as the worstcase false alarm
rate goes to 0. The insights from these theoretical results are corroborated by
a simulation study.

In this paper, the problem of quickly detecting an abrupt change on a
stochastic process under Bayesian framework is considered. Different from the
classic Bayesian quickest changepoint detection problem, this paper considers
the case where there is uncertainty about the postchange distribution.
Specifically, the observer only knows that the postchange distribution belongs
to a parametric distribution family but he does not know the true value of the
postchange parameter. In this scenario, we propose two multichart detection
procedures, termed as MSR procedure and modified MSR procedure respectively,
and show that these two procedures are asymptotically optimal when the
postchange parameter belongs to a finite set and are asymptotically
$\epsilon$optimal when the postchange parameter belongs to a compact set with
finite measure. Both algorithms can be calculated efficiently as their
detection statistics can be updated recursively. We then extend the study to
consider the multisource monitoring problem with unknown postchange
parameters. When those monitored sources are mutually independent, we propose a
windowbased modified MSR detection procedure and show that the proposed
detection method is firstorder asymptotically optimal when postchange
parameters belong to finite sets. We show that both computation and space
complexities of the proposed algorithm increase only linearly with respect to
the number of sources.

We analyze an optimal stopping problem with a constraint on the expected
cost. When the reward function and cost function are Lipschitz continuous in
state variable, we show that the value of such an optimal stopping problem is a
continuous function in current state and in budget level. Then we derive a
dynamic programming principle (DPP) for the value function in which the
conditional expected cost acts as an additional state process. As the optimal
stopping problem with expectation constraint can be transformed to a stochastic
optimization problem with supermartingale controls, we explore a second DPP of
the value function and thus resolve an open question recently raised in [S.
Ankirchner, M. Klein, and T. Kruse, A verification theorem for optimal stopping
problems with expectation constraints, Appl. Math. Optim., 2017, pp. 133].
Based on these two DPPs, we characterize the value function as a viscosity
solution to the related fully nonlinear parabolic HamiltonJacobiBellman
equation.

We solve the martingale optimal transport problem for cost functionals
represented by optimal stopping problems. The measurevalued martingale
approach developed in ArXiv: 1507.02651 allows us to obtain an equivalent
infinitedimensional controllerstopper problem. We use the stochastic Perron's
method and characterize the finite dimensional approximation as a viscosity
solution to the corresponding HJB equation. It turns out that this solution is
the concave envelope of the cost function with respect to the atoms of the
terminal law. We demonstrate the results by finding explicit solutions for a
class of cost functions.

We solve the problem of optimal stopping of a Brownian motion subject to the
constraint that the stopping time's distribution is a given measure consisting
of finitelymany atoms. In particular, we show that this problem can be
converted to a finite sequence of stateconstrained optimal control problems
with additional states corresponding to the conditional probability of stopping
at each possible terminal time. The proof of this correspondence relies on a
new variation of the dynamic programming principle for stateconstrained
problems which avoids measurable selection. We emphasize that distribution
constraints lead to novel and interesting mathematical problems on their own,
but also demonstrate an application in mathematical finance to modelfree
superhedging with an outlook on volatility.

We consider the superhedging price of an American option in a discretetime
market in which stocks are available for dynamic trading and European options
are available for static trading. We show that the superhedging price $\pi$ is
given by the supremum over the prices of the American option under randomized
models. That is, $\pi=\sup_{(c_i,Q_i)_i}\sum_ic_i\phi^{Q_i}$, where $c_i \in
\mathbb{R}_+$ and the martingale measure $Q^i$ are chosen such that $\sum_i
c_i=1$ and $\sum_i c_iQ_i$ prices the European options correctly, and
$\phi^{Q_i}$ is the price of the American option under the model $Q_i$. Our
result generalizes the example given in ArXiv:1604.02274 that the highest model
based price can be considered as a randomization over models.

How do largescale participants in parimutuel wagering events affect the
house and ordinary bettors? A standard narrative suggests that they may
temporarily benefit the former at the expense of the latter. To approach this
problem, we begin by developing a model based on the theory of large
generalized games. Constrained only by their budgets, a continuum of diffuse
(ordinary) players and a single atomic (largescale) player simultaneously
wager to maximize their expected profits according to their individual beliefs.
Our main theoretical result gives necessary and sufficient conditions for the
existence and uniqueness of a purestrategy Nash equilibrium. Using this
framework, we analyze our question in concrete scenarios. First, we study a
situation in which both predicted effects are observed. Neither is always
observed in our remaining examples, suggesting the need for a more nuanced view
of largescale participants.

On a filtered probability space
$(\Omega,\mathcal{F},P,\mathbb{F}=(\mathcal{F}_t)_{t=0,\dotso,T})$, we consider
stopperstopper games $\overline
V:=\inf_{\Rho\in\bT^{ii}}\sup_{\tau\in\T}\E[U(\Rho(\tau),\tau)]$ and
$\underline V:=\sup_{\Tau\in\bT^i}\inf_{\rho\in\T}\E[U(\Rho(\tau),\tau)]$ in
discrete time, where $U(s,t)$ is $\mathcal{F}_{s\vee t}$measurable instead of
$\mathcal{F}_{s\wedge t}$measurable as is often assumed in the literature,
$\T$ is the set of stopping times, and $\bT^i$ and $\bT^{ii}$ are sets of
mappings from $\T$ to $\T$ satisfying certain nonanticipativity conditions. We
convert the problems into a corresponding Dynkin game, and show that $\overline
V=\underline V=V$, where $V$ is the value of the Dynkin game. We also get the
optimal $\Rho\in\bT^{ii}$ and $\Tau\in\bT^i$ for $\overline V$ and $\underline
V$ respectively.

In this paper we develop two numerical methods for optimal stopping in the
framework of one dimensional diffusion. Both of the methods use the Skorohod
embedding in order to construct recombining tree approximations for diffusions
with general coefficients. This technique allows us to determine convergence
rates and construct nearly optimal stopping times which are optimal at the same
rate. Finally, we demonstrate the efficiency of our schemes on several models

We analyze the asymptotic behavior for a system of fully nonlinear parabolic
and elliptic quasi variational inequalities. These equations are related to
robust switching control problems introduced in [3]. We prove that, as time
horizon goes to infinity (resp. discount factor goes to zero) the long run
average solution to the parabolic system (resp. the limiting discounted
solution to the elliptic system) is characterized by a solution of a nonlinear
system of ergodic variational inequalities. Our results hold under a
dissipativity condition and without any non degeneracy assumption on the
diffusion term. Our approach uses mainly probabilistic arguments and in
particular a dual randomized game representation for the solution to the system
of variational inequalities.

This paper studies the market viability with proportional transaction costs.
Instead of requiring the existence of strictly consistent price systems (SCPS)
as in the literature, we show that strictly consistent local martingale systems
(SCLMS) can successfully serve as the dual elements such that the market
viability can be verified. We introduce two weaker notions of no arbitrage
conditions on market models named no unbounded profit with bounded risk (NUPBR)
and no local arbitrage with bounded portfolios (NLABP). In particular, we show
that the NUPBR and NLABP conditions in the robust sense for the smaller bidask
spreads is the equivalent characterization of the existence of SCLMS for
general market models. We also discuss the implications for the utility
maximization problem.