Fx Asian Option Pricing using Quantum Computers

Rafał Pracht1 , Adam Ryterski1 , Julia Plewa, and Marek Stefaniak1

BNP Paribas, ul. Kasprzaka 2; 01-211 Warszawa

{rafal.pracht,adam.ryterski,marek.stefaniak}@bnpparibas.pl

Abstract. This work presents an algorithm for calculating Asian-style currency options using
a gate-base quantum computer. We have described the algorithm in detail, starting from
loading the distribution of the underlying asset, through the comparator, calculating the
average, payout function, and ending on QAE. We show a novel approach in distribution
loading which allow extending our solution to a local volatility model (the previous papers
use constant values for interest rate and volatility for the underlying asset). We compared
the results obtained on the quantum computer with those obtained classically for different
strikes and conducted an error analysis. This work provides a starting point to further work
on the deployment Quantum Monte Carlo method in a financial institution.

Keywords: Quantum Monte Carlo · Pricing derivatives · Quantum Computing

1 Introduction
FX options are derivative instruments used to manage currency-linked risks. There are many dif-
ferent types of currency options available in the OTC market. All of them offer various degrees of
participation in favorable exchange rate movements and are an attractive alternative to simple FX
instruments.
An important class of FX options is Asian options, which are contracts with payoffs depending on
the average exchange rate over some time period. Such instruments are commonly used by corporate
treasurers, as they are often lower priced and better suited for hedging risk related to cash flows
exchanged on a regular basis. The main drawback of these contracts is that they are usually hard
to price. There is no closed-form pricing formula and even though the industry developed many
numerical pricing methods, the classical Monte Carlo algorithms are still very often used for valuing
these contracts.
Recent developments in the field of quantum computing provide instruments that allow for
a theoretical quadratic speed-up compared to classical Monte Carlo methods. In this article, we
investigate the possible application of quantum computing in option pricing by developing a Quan-
tum Monte Carlo algorithm for valuing Asian options in Black-Scholes-Merton model with time-
dependent parameters.
In the opening chapter, we will present a mathematical model of the underlying asset price
changes and we will also review both classical and quantum algorithms for option pricing.
Later we will describe in detail the Asian option pricing algorithm on a gate-based quantum
computer. We’ll start with a general overview of the algorithm. Then we describe in detail how to
load the EUR / PLN prices onto a quantum computer and how this algorithm could be extended to
use the local variance model, which is one of the most important models used on financial markets.
Next, we describe how you can calculate the arithmetic mean without using auxiliary qubits. Later,
we will describe the algorithm for calculating the option payoff function. We end this chapter with

Electronic copy available at: https://ssrn.com/abstract=4137397

2 R. Pracht et al.

the description of the QAE algorithm and the description of the postprocessing that should be
performed on a classic computer in order to determine the value of the option.
In the next chapter, we will present the results of the Asian option valuation using our algorithm
and compare them with the results obtained with the classic Monte Carlo.
In the last chapter, we will present our conclusions and further directions of our research work.

2 Review of option pricing algorithms

In this section, we describe the option pricing problem and discuss the current state of research
in that field, including both the classical state-of-the-art algorithms, as well as the more recent
quantum approaches.

2.1 The mathematical description of the problem

We narrow our research to the problem of pricing European-style Asian options in Black-Scholes-
Merton model with time-dependent parameters. The basic model was introduced in [4,25] and a
detailed description of the extended model can be found for example in [32,7]. An overview of other
option types and FX market conventions can be found in the book by Wystup [46].
In our analysis, we are referring only to the FX options, but the methods described can be also
used for other underlying assets. We assume later in the article that a spot exchange rate St is
quoted in FOR-DOM convention, which means that one unit of the foreign currency (FOR) costs
St amount of domestic currency (DOM).

Types of options A European plain vanilla option is a financial instrument that gives the holder
the right (but not the obligation) to exchange a given currency at a predetermined price (strike
price) at a future date (expiration date)1 . A buyer of the contract acquires protection against
adverse exchange rate movements and in return pays a premium for the option. The payoff function
of such contract is strategy dependent, i.e.:

N f · max(ST − K, 0) for call option, (1)

N f · max(K − ST , 0) for put option, (2)

where N f denotes the nominal of the option in FOR currency, T is the expiration date, ST is the
spot exchange rate at expiration date and K is the strike price.
An Asian option also known as the Average rate option is a contract written on the average of
the exchange rate over a predefined period of time. Such a contract provides the buyer protection
against the risk of adverse currency movements in the average exchange rate over the period of time
defined in contract terms. The payoff function of such an option is determined by the path taken
by the exchange rate and can be expressed as follows:

N f · max(AT − K, 0) for call option, (3)

f
N · max(K − AT , 0) for put option, (4)
1
The contract can be either cash or physically settled.

Electronic copy available at: https://ssrn.com/abstract=4137397

 Fx Asian Option Pricing using Quantum Computers 3

where AT is the average exchange rate realized over a predefined period of time. The average rate
can be specified in a number of ways and over various time periods. Contracts traded in the market
are always cash settled and usually based on discrete arithmetic average rate of the form:
n
1X
AT = St . (5)
n i=1 i

where prices are measured at the time points 0 ≤ t1 ≤ t2 ≤ . . . ≤ tn ≤ T .

Valuation in risk-neutral world Lets assume that (Ω, F, F, Q) is filtered probability space with
domestic risk-neutral measure Q under which the price process S = (St )t∈[0,T ] follows the stochastic
differential equation (SDE):
dSt
= (rtd − rtf )dt + σt dWt , (6)
St
where the domestic and foreign instantaneous risk-free rates rd and rf are deterministic functions
of time. The W is a one-dimensional F-adapted standard Brownian motion representing the ran-
domness of the price process with amplitude measured by the instantaneous volatility σt which is
also a deterministic function of time.
The model parameters are usually calibrated to the term structure of forward rates and option
volatilities observed in the market. The model value of the European-style option with payoff
function H can be computed as the expected payoff value under the risk-neutral probability measure,
i.e.: RT d
V0 = e− 0 r (t)dt · EQ [H(S)] (7)
In the next section we review the most common approaches used to calculate option values in
practice.

2.2 Classical algorithms for Fx Asian option pricing

There is no closed-form solution for the value of Arithmetic average rate option in Black-Scholes-
Merton model, however a number of numerical pricing techniques have been introduced in the
literature. One of such methods is Monte Carlo simulation which was introduced by Boyle in [5]. In
this approach we simulate the risk-neutral price process multiple times using appropriate numerical
scheme for SDE 6. In this paper we use Euler-Maruyama scheme of the form:

Stn+1 = Stn + (rtdn − rtfn )Stn (tn+1 − tn ) + σtn Stn ∆Wn (8)

where ∆Wn = Wtn+1 − Wtn and 0 ≤ t1 ≤ t2 ≤ . . . ≤ tn+1 ≤ T is a predefined discretization of time

space. At the next step of the algorithm we compute the payoffs of the option at each realization
of the process and approximate the expectation 7 with the discounted average value of simulated
payoffs, i.e.
PM
RT d H(S k )
V̂0M = e− 0 r (t)dt · k=1 (9)
M
where ∀t∈[0,T ] (Stk )k=1,2,...,M is a sequence of realizations of risk-neutral price process S and H is
the payoff function of the option.

Electronic copy available at: https://ssrn.com/abstract=4137397

4 R. Pracht et al.

The Law ofpLarge Numbers and the Central Limit Theorem provide that V̂M converges to V
with speed O( var(H(S))/M ). Hence we conclude that the accuracy of the estimate shrinks like
the inverse square root of the number of samples and it is proportional to the standard deviation of
H(S). The former is the largest drawback of the Monte Carlo methods, as the satisfying precision
of the estimate involves generating a large amount of simulated price paths, which is usually time
consuming. On the other hand the later allows for application of variance reduction techniques as
it was presented by Kemna and Vorst in [22].
An alternative approach is to approximate the unknown distribution of the arithmetic mean
price with lognormal distribution as it was presented by Levy in [24] or Turnbull and Wakeman
in [38]. Later Milevsky and Posner [27] showed that it is more intuitive to use a reciprocal gamma
distribution instead of lognormal distribution, as under suitable parameter restrictions the infinite
sum of correlated lognormal random variables is reciprocal gamma distributed.
Another way to approximate Arithmetic average option price is to use the formula for Geometric
average option price. Based on the relationships between arithmetic and geometric averages Vorst
[41] derives upper and lower bound for the Arithmetic average option price and by adjusting the
strike price introduce approximation formula for the option value. Curran [10] on the other hand
presents alternative approximations derived by conditioning on the geometric mean price.
A different class of algorithms used for pricing Asian options are lattice methods, which are
modifications of the binomial Cox, Ross and Rubinstein model described in [8]. Such approach
was proposed by Hull and White in [19] and later improved by Forsyth, Vetzal and Zvan [14].
In this case the binomial model was augmented by an auxiliary state vector of averages at each
node. As the set of possible average values grow exponentially with the number of nodes in the
lattice, so a representative set of average values was chosen and interpolation was used to calculate
corresponding option values in backward in time pricing procedure.
The price of an Asian option can be computed by numerically solving a partial differential
equation (PDE) in two-dimensional space, as it was presented by Ingersoll [20] or Wilmot at al.
[44]. Rogers and Shi [30] further reduce the valuation problem to a one-dimensional PDE, however it
was difficult to solve numerically. An alternative reduction to one-dimensional PDE was suggested
by Večeř [40], who studied the connection between Asian contracts and options on traded accounts.

2.3 Quantum algorithms for option pricing

Classical Monte Carlo algorithm is widely used in pricing derivatives, but in many advanced instru-
ments it is required to make a huge amount of simulation to get a price with acceptable error. In
practice, it may turn out to be an expensive and slow method especially for institutions that have
to price large amounts of instruments.
Patrick Rebentrost in [29] proposed a quantum algorithm to price European options with
quadratic speed-up over classical methods. The algorithm consists of 3 steps:
– Loading probability distribution into quantum state. Grover and Rudolph in [18] shows that it
is possible to create a operator to load any efficiently integrable probability distribution in the
quantum state. The article [47] propose how to load generic probability distribution given by
data samples using quantum Generative Adversarial Networks (qGANs).
– Calculating payoff. To define the payoff function described in 1 we need to use an additional
qubit (ancilla qubit) to store the state of expectation value. Payoff calculation is based on
controlled rotation ancilla qubit and calculation the probability of measuring ancilla qubit in
state |1⟩

Electronic copy available at: https://ssrn.com/abstract=4137397

 Fx Asian Option Pricing using Quantum Computers 5

– Estimating probability. To know the probability of measuring an ancilla qubit in state |1⟩ it
can be using Quantum Phase Estimation algorithm to estimate this value. It is important to
notice, that QPE provides quadratic speed-up over classical Monte Carlo method with the same
calculation error.

3 Implementation on a gate based quantum computer

In this section, we will quote the Asian currency option using a quantum algorithm. The underlying
asset is the exchange rate of the EUR/PLN currency, modeled by a stochastic differential equation
described in Equation 6. We assume four time periods shall be used to calculate the mean. Below
you can find the option and underlying model parameters:

– j ∈ {1, . . . , M },
90 182 274 365
– tj = [ 365 , 365 , 365 , 365 ],
d
– rtj = [2.79%, 3.383%, 3.596%, 3.661%],
– rtfj = [−0.61%, −0.551%, −0.491%, −0.430%],
– σtj = [6.036%, 6.275%, 6.146%, 6.607%],
– Strike = 4.6265,
– the current exchange rate S0 = 4.6265.

The entire algorithm consists of the following steps:

– Encoding the option value in the probability of measuring |1⟩ in ancilla qubit |p⟩. This part of
algorithm consist the following steps, the overview of them can be show on the Figure 12 .
• Distribution loading
• Calculation of mean
• Compare
• Payoff function
– Next, we have to extract the probability |1⟩ from |p⟩ qubit, see Section 3.5.
– In the last step we extract the option price from probability, see Section 3.6.

2
All symbols used will be described further in the paper.

Electronic copy available at: https://ssrn.com/abstract=4137397

6 R. Pracht et al.

|0⟩N |I1 ⟩N

|0⟩N |I2 ⟩N
Px
.. ..
. .

|0⟩N |IM ⟩N

1
PM
|0⟩Ñ M j=1 Ij |S⟩Ñ

|0⟩1 S ≥ Strike |c⟩1

|0⟩1 max(S − Strike, 0) |p⟩1

Fig. 1: The quantum circuit to encode the option price in ancilla qubit.

3.1 Distribution loading

The first stage of the algorithm is to load into the quantum register the probability distribution
generated by Equation 6. We assume M = 4 points tj indexed by j ∈ {1, . . . , M }, shall be used to
calculate the mean and hence the option price. Let Stj be the EUR/PLN exchange rate value at
the time point tj . For each of the points, we use exactly the same discretization, which maps Stj
to an appropriate interval index Ij . This allows us to simplify the calculation of the mean and has
no notable impact on the discretization error. The minimum and maximum values of the exchange
rate (labeled here as low and high) should be chosen carefully. Too large of a range increases
the discretization error, while a small range can also have a negative impact on the estimation,
because it excludes the exchange rates outside this interval. To choose the good values of the ends
of the compartment, we use as a guideline the mean and standard deviation of the underlying asset
distribution in the maturity. We would like to cover 6 standard deviation in our distribution interval
therefore we use the following definition of the high and low: low = µ − 3σ and high = µ + 3σ . In
our case, we use low = 3.941 and high = 5.949.
For each of the time points, we use N = 3 qubits to store the probability distribution. Given
the number of qubits N , we have 2N intervals of the exchange rate values. Increasing the number
of qubits decrease the discretization error, so we should use the largest number as allowed by the
limitations of the current quantum machines. Let’s define the distance between the middle points
of two adjacent intervals as
high − low
∆= (10)
2N − 1
Using the low value of the exchange rate and ∆, we can define the following intervals:

∀j∈{1,...,M } , {0, . . . , 2N − 1} ∋ Ij
∆ ∆ ∆ ∆ (11)
7→ [low + ∆ ∗ Ij − , low + ∆ ∗ Ij + ] ∈ [low − , high + ]
2 2 2 2

Electronic copy available at: https://ssrn.com/abstract=4137397

 Fx Asian Option Pricing using Quantum Computers 7

For each of the intervals, we use the mean value as the exchange rate used in option pricing
described in Equation 5. The mapping from qubit index to the EUR/PLN value is defined as

high − low
∀j∈{1,...,M } , {0, . . . , 2N − 1} ∋ Ij 7→ ∗ It + low = ∆ ∗ Ij + low ∈ [low, high] (12)
2N − 1
Next, our goal is to load the probability distribution on a quantum computer and calculate the
state ψ. By this, we mean creating a superposition of all possible paths of price changes of the
√
underlying |I1 ⟩N |I2 ⟩N . . . |IM ⟩N with weights expressing the probability of such path pI1 ,I2 ,...,IM .
N N N
2X −1 2X −1 2X −1
√
|0⟩M ∗N 7→ |ψ⟩M ∗N = ... pI1 ,I2 ,...,IM |I1 ⟩N |I2 ⟩N . . . |IM ⟩N (13)
I1 =0 I2 =0 IM =0

where pI1 ,I2 ,...,IM is the probability of the exchange rate on the path I1 , I2 , . . ., IM
The method of generating the EUR/PLN exchange rate paths and their probability can be found
in Algorithm 2.

Algorithm 1 The generation of the PDF of the Black-Scholes-Merton model.

function generate_pdf(S, r, σ, ∆t)
µ ← e(r−0.5σ)∆t+ln(S)
√
σ ← σ ∆t
return pdf(x) ← lognorm.pdf(x, µ, σ)
end function

After preparing the probabilities for all paths, we load them into the quantum computer using
the method described in [21] using its Qiskit implementation [2].
The marginal probability of the EUR/PLN exchange rate on the maturity calculated according
to Equation 14 can be found in Figure 2. In Figure 3 you can find the same distributed but with
mapping to qubits values.
N N N
2X −1 2X −1 2X −1
√
... pI1 ,I2 ,...,IM |IM ⟩N (14)
I1 =0 I2 =0 IM −1 =0

It is worth noting that in contrast to the method described in [33] which makes the assumption
that the interest rate and the volatility are constant across all time periods, we use the local volatility
model described in [13]. This allows us to include the volatility smile in our pricing method.

Electronic copy available at: https://ssrn.com/abstract=4137397

8 R. Pracht et al.

Algorithm 2 The generation of the EUR/PLN exchange rate paths and their probability.
function BSM_distribution(S0 , ⃗r, ⃗σ , ∆t, ⃗ low, N )
S ← S0
pdf1 (x) ← GENERATE_PDF(S0 , r0 , σ0 , ∆t0 ) ▷ We use Algorithm 1
for I1 ← 0 to 2N − 1 do
S1 ← ∆ ∗ I1 + low ▷ We use Equation 12
p1 ← pdf1 (S1 )
pdfS2 (x) ← GENERATE_PDF(S1 , r1 , σ1 , ∆t1 )
for I2 ← 0 to 2N − 1 do
S2 ← ∆ ∗ I2 + low
p2 ← pdfS2 (S2 )
pdf3 (x) ← GENERATE_PDF(S2 , r2 , σ2 , ∆t2 )
for I3 ← 0 to 2N − 1 do
S3 ← ∆ ∗ I3 + low
p3 ← pdf3 (S3 )
pdf4 (x) ← GENERATE_PDF(S3 , r3 , σ3 , ∆t3 )
for I4 ← 0 to 2N − 1 do
S4 ← ∆ ∗ I4 + low
p4 ← pdf4 (S4 )
pI1 ,I2 ,I3 ,I4 ← p1 p2 p3 p4
push [S1 , S2 , S3 , S4 ] on S
push pI1 ,I2 ,I3 ,I4 on p
end for
end for
end for
end for
return S = [[S1 , S2 , S3 , S4 ], . . .], p = [[pI1 ,I2 ,I3 ,I4 ], . . .]
end function

Fig. 2: EUR/PLN exchange rate. Fig. 3: The qubits mapping.

Electronic copy available at: https://ssrn.com/abstract=4137397

 Fx Asian Option Pricing using Quantum Computers 9

3.2 Calculation of mean

The next stage is to calculate the mean of the price of the underlying on the given path using the
Equation 5. It is necessary to note that to store the M numbers encoded on N bits we need log2 M
additional bits to store the result. So we introduce a new symbol Ñ which means the number of
qubits required to store the mean of the exchange rates. Let denote Ñ = N + log2 M , now our goal
is to do the transformation on Equation 15 which calculates the mean of exchange rates and store
the result in the quantum register with Ñ qubits.

N N N
2X −1 2X −1 2X −1 M
√ X √ 1 X
... pI1 ,I2 ,...,IM |I1 ⟩N |I2 ⟩N . . . |IM ⟩N 7→ pI1 ,I2 ,...,IM | It ⟩
M t=1
I1 =0 I2 =0 IM =0 I1 ,...,IM N +log2 M

X √ M
1 X
7→ pI1 ,I2 ,...,IM | It ⟩
M t=1
I1 ,...,IM Ñ
(15)

There are many algorithms that implement the arithmetic operations on quantum computer.
Some of them are reversible implementation of the classical approach [3,39], some expand on the
classical approach with custom modifications and improvements [15,9,12,35,36], while other algo-
rithms are based on a more quantum-specific design [26,37,43]. In our case, we will use an algorithm
based on the Quantum Fourier Transform (QFT) proposed by Draper in [11] and further described
in [31]. The main reason to use QFT-based algorithm is the elimination of the need to use additional
ancillary qubits.
We would like to add together the index values I1 , I2 , . . . , IM of the currency exchange rate. Note
that It is an integer value, which means that we can write it as the following binary representation
It = It,1 2N −1 + It,2 2N −2 + ... + It,N 20 . The value is then encoded as the binary representation in
qubits |It ⟩ = |It,1 ⟩⊗|It,2 ⟩⊗...⊗|It,N ⟩. The procedure of preparing this representation was described
in Section 3.1.
First, we calculate the Quantum Fourier transform of the last register IM extended by log2 M
qubits (in our case 2), evolving |IM ⟩N +log2 M = |IM ⟩Ñ into ϕ(IM )N +log2 M = ϕ(IM )Ñ . Formally,
the Quantum Fourier transform is defined as

N −1
1 X i 2πxk
|ϕ(x)⟩ = QFT |x⟩ = e N |k⟩ . (16)
N
k=0

The Inverse Quantum Fourier Transform (IQFT) is defined as

d−1
1 X −i 2πxk
IQFT |k⟩ = √ e d |x⟩ . (17)
d x=0

It is worth noting that by applying IQFT after QFT we return to the initial state:

IQFT |ϕ(x)⟩ = IQFT QFT |x⟩ = QFT−1 QFT |x⟩ = |x⟩ . (18)

Electronic copy available at: https://ssrn.com/abstract=4137397

10 R. Pracht et al.

Therefore, after the transformation, we have following state

N N N
2X −1 2X −1 2X −1
√
... pI1 ,I2 ,...,IM |I1 ⟩N |I2 ⟩N . . . |IM ⟩N +log2 M
I1 =0 I2 =0 IM =0
N N N
2X −1 2X −1 2X −1
√
7→ ... pI1 ,I2 ,...,IM |I1 ⟩N |I2 ⟩N . . . QFT |IM ⟩Ñ (19)
I1 =0 I2 =0 IM =0
N N N
2X −1 2X −1 2X −1
√
7→ ... pI1 ,I2 ,...,IM |I1 ⟩N |I2 ⟩N . . . |ϕ(IM )⟩Ñ .
I1 =0 I2 =0 IM =0

Once we have |ϕ(IM )⟩Ñ , we can add the numbers

|I1 ⟩N . . . |IM −1 ⟩N using controlled phase rotation gates as in Draper’s scheme [31], so we perform
the following operation:

(IM,j +It,j )2Ñ −j ks 2Ñ +1−s (IM,j +It,j )ks

{0, . . . , M − 1} ∋ t 7→ e
2πi
2Ñ +1 =e
2πi
2j+s−Ñ
(20)

that depends on jth qubits of the representation of the numbers to be added and is applied on the
sth qubit in transformed register [31].
The procedure described above performs the modular addition arithmetic, but we need to allo-
cate the proper size for the result register or otherwise overflow issues could occur. Therefore, we
use Ñ qubits instead of N to store the final result. Equation 20 is implemented by the circuit shown
in Figure 4 and by Algorithm 3.

|I1,1 ⟩

..
.

|I1,N ⟩ = |I1,3 ⟩

|IM −1,1 ⟩ = |I3,1 ⟩

..
.

|IM −1,N ⟩ = |I3,3 ⟩

π π
|ϕ(IM,1 )⟩ = |ϕ(I4,1 )⟩

..
.
π π π π
|ϕ(IM,Ñ )⟩ = |ϕ(I4,5 )⟩ 16 8 16 8

Fig. 4: The addition circuit (in the transformation domain)

Electronic copy available at: https://ssrn.com/abstract=4137397

 Fx Asian Option Pricing using Quantum Computers 11

Algorithm 3 QFT-based addition.

Ensure: |I1 ⟩N |I2 ⟩N . . . |IM ⟩Ñ ]
QF T (|ϕ(IM )⟩Ñ , do_swaps = F alse) ▷ We use Equation 16
for t ← 1 to M − 1 do
for j ← 0 to N − 1 do
idx ← 0
for u ← j to Ñ − 1 do
2π
cp( 2idx+1 , |It,j ⟩ , |IM,u ⟩)
idx ← idx + 1
end for
end for
end for
IQF T (|ϕ(IM )⟩Ñ , do_swaps = F alse) ▷ We use Equation 17

It is worth noting that for performance reasons we don’t swap QFT and IQFT but instead of
this we remap the indexes in the inner loops of Algorithm 3.
The last and the easy step is to divide the sum by M time steps. The final sum has the following
binary representation S = S1 2N −1 + S2 2N −2 + ... + SN 20 , and in our case, the number of the time
periods can be written as M = 4 = 22 . We only need to treat the sum as a fixed-point number with
2 qubits on the fractional part, as shown in the following equation:

S S1 2N −1 + S2 2N −2 + ... + SN 20
= = S1 2N −3 + S2 2N −4 + ... + SN 2−2 (21)
M 22
So we map the sum to the EUR/PLN exchange rate in the following way:
∆
∆˜ = , {0, . . . , 2Ñ −1 } ∋ j 7→ low + ∆˜ ∗ j 7→ Sj (22)
M
The state after our transformation is described by
Ñ −1
2X
√
pj |Sj ⟩Ñ , Sj = low + ∆˜ ∗ j. (23)
j=0

3.3 Compare
The next stage is to flag the ancilla qubit whenever the value of the EUR/PLN exchange rate is
greater or equal to the option strike. So we use a quantum comparator circuit that sets a temporary
qubit |c⟩, initially in state |0⟩ to state |1⟩, if S ≥ Strike = K and |0⟩ otherwise [33]. Our goal is to
perform the following transformation:
Ñ −1
2X
√ X√ X√
pj |Sj ⟩Ñ |0⟩1 7→ pj |Sj ⟩Ñ |0⟩1 + pj |Sj ⟩Ñ |1⟩1 (24)
j=0 j<K j≥K

There are many algorithms to do that, we use the one proposed in [33] and based on the
ripple-carry addition circuit described in [9]. The other option is to use the algorithm described

Electronic copy available at: https://ssrn.com/abstract=4137397

12 R. Pracht et al.

in the ’Constraint testing’ chapter of the paper [16] or use the adder exploiting Quantum Fourier
Transform described in [11] and [31].
Because we know the strike price ahead of computation, we can tailor our circuit to a concrete
value. The main idea of the algorithm is to add the negative number U 2(K) to the price register S
and check if there is a carry. The algorithm requires the Ñ ancilla qubits initialized as anc = |0⟩Ñ ,
and it consists of the followed steps:

1. Convert the strike price Strike = K to a binary number.

2. Calculate the two’s complement of the binary number and store the result in an array k.
3. The carry from the first bit of S can be only when S1 is equal to 1 and k[1] is equal to 1.
To check this we add cx gate controlled by S1 and with anc[1] as a target, if k[1] = 1 and do
nothing in the other case.
4. For the remaining bits j, we do as follow:
– if k[j] = 0, The carry can be only when the last ancilla qubit anc[j − 1] is equal to 1 and
also the Sj = 1. This can be performed by a single Toffoli gate.
– if k[j] = 1, The carry can be only when Sj = 1 ∨ anc[j − 1] = 1. This can be performed by
an OR gate, which can be decomposed to a Toffoli gate and 5 X gates.

The algorithm can be also found in Figure 5.

3.4 Payoff function

The next stage of the algorithm requires encoding the option’s value in the probability of measuring
a 1 in the result qubit. The method of calculating option prices for European and Asian style options
was presented by Stefan Woerner and Daniel J. Egger [45] and Nikitas Stamatopoulos et al. [33].
For a call option, its value on maturity can be calculated using the equation Pcall = max(ST −
K, 0). Our goal is to create a quantum operator which encodes in the state |1⟩ ancilla qubit the
square root of the option price for a given value of the EUR/PLN exchange rate.
p p 
|Sj ⟩Ñ |0⟩ 7→ |Sj ⟩Ñ 1 − fj |0⟩ + fj |1⟩ , (25)

where f (j) is defined as

( ( (
0 j<K 0 j<K 0 j<K
f (j) = = = (26)
Sj − Strikej j≥K ˜
low − Strike + ∆j j≥K ˜
fmin + ∆j j ≥ K.

By combining Equations 23 and 25, we arrive at

Ñ −1 Ñ −1
2X 2X
√ √ p p 
pj |Sj ⟩Ñ |0⟩ 7→ pj |Sj ⟩Ñ 1 − fj |0⟩ + fj |1⟩ (27)
j=0 j=0

And by skipping the mean of the EUR/PLN exchange rate in Equation 27, we arrive at
Ñ −1 Ñ −1
2X 2X
√ p √ p
pj 1 − fj |0⟩ + pj fj |1⟩ (28)
j=0 j=0

Electronic copy available at: https://ssrn.com/abstract=4137397

 Fx Asian Option Pricing using Quantum Computers 13

k[1] = 1
|S1 ⟩

|S2 ⟩ X X

..
. k[Ñ] = 0 k[Ñ] = 1

|SÑ ⟩ X X

|anc1 ⟩ X X

|anc2 ⟩ X

.. k[2] = 0 k[2] = 1
.

|ancÑ −1 ⟩ X X

|ancÑ ⟩ X

|c⟩

Fig. 5: The compare circuit

Electronic copy available at: https://ssrn.com/abstract=4137397

14 R. Pracht et al.

So the probability of measuring |1⟩ in the ancilla qubit in Equation 28 is equal to

Ñ −1 Ñ −1 Ñ −1
2X √ p 2 2X 2X
P1 = pj fj = pj fj = pj max(Sj − K, 0) = E[Pcall ]. (29)
j=0 j=0 j=0

To correctly defined function f (j) we need some tricks, first we need a function with domain
[−1, 1] we do remap the function f (j) and define a new one f˜(i) ∈ [−1, 1] using the following
equation:
2f (j) 2f (j)
f˜(j) = −1= −1 (30)
Ñ ˜
fmin + (2 − 1) ∗ ∆ fmax

The required domain [−1, 1] of our function f˜ is needed to utilize the sine function property.
Namely, we utilize the fact that sin2 (x + π4 ) ≈ x + 21 for x values near zero, in our case we have
sin2 (cf˜(i) + π4 ) ≈ cf˜(i) + 12 for small values of c.
Let’s re-write the function f˜ as
(
˜ π g0 j<K
cf (j) + = (31)
4 g0 + g(j) j ≥ K

So we have to find the constant g0 and the function g(j) which satisfies Equations 26, 30, and
31. It is easy to show that it is required for g0 and g(j) to be equal to
π
g0 = − c,
4
(32)
!
fmin ∆˜
g(j) = 2c + j .
fmax fmax

We can easily implement this transformation on gate-based quantum computers using the control
Ry gate. This implementation can be seen in Algorithm 4 and in the circuit shown in Figure 6.

Algorithm 4 The algorithm to calculate the pay-off function of the call option.
flow = low − Strike
˜
fmax = flow + (2Ñ − 1) ∗ ∆
g0 = π4 − c ▷ We use Equation 32
gj,0 = 2cf low
fmax
˜
gj = f2c∆
max
ry(−2 · g0 , |payoff⟩)
cry(−2 · gj,0 , |c⟩ , |payoff⟩)
for k ← 0 to Ñ − 1 do
ccry(−2 · 2k · gj , |S0,...,k,...,Ñ −1 ⟩ , |c⟩ , |payoff⟩)
end for

Electronic copy available at: https://ssrn.com/abstract=4137397

 Fx Asian Option Pricing using Quantum Computers 15

|S1 ⟩

|S2 ⟩

..
.

|SÑ ⟩

|c⟩

|payoff⟩ Ry(−2g0 ) Ry(−2gj,0 ) Ry(−20 gj ) Ry(20 gj ) Ry(−21 gj ) Ry(21 gj ) Ry(−2Ñ gj ) Ry(2Ñ gj )

Fig. 6: The pay-off function circuit.

3.5 Amplitude Estimation

After performing the steps described in the previous sections, the second to last step is to retrieve
the probability of the result qubit being in the |1⟩ state, since our solution is encoded in this
probability. In some of our initial experiments, we achieved this by relying on Qiskit’s statevector
simulator [2], but for real-life applications this requires the use of amplitude estimation algorithms.
Quantum Amplitude Estimation (QAE) was first described by Brassard et al. in 2000 [6]. Given
the operator A such that
√ √
A |0⟩n |0⟩ = 1 − a |ψ0 ⟩n |0⟩ + a |ψ1 ⟩n |1⟩ , (33)
where a is the unknown probability and |ψ0 ⟩n and |ψ1 ⟩n are two normalized states, the algorithm
estimates the value of a with such probability that the error is inversely proportional to the number
of applications of A. QAE follows the same template as Quantum Phase Estimation (QPE) [23],
but it replaces the controlled U rotations used in QPE with the Grover operator, which is defined
as
Q = AS0 A† Sψ0 , (34)
where S0 and Sψ0 are rotations about the states |0⟩ and |ψ0 ⟩, as defined in
Sψ0 = I − 2 |ψ0 ⟩ ⟨ψ0 | ⊗ |0⟩ ⟨0| ,
(35)
S0 = I − 2 |0⟩n+1 ⟨0|n+1 .
The main downside of QAE is that it’s computationally expensive. Additionally, it requires a
number of auxiliary qubits that determine the final accuracy of the results and the only way to
improve the results is to add more qubits. In recent years, many variants of QPE-free amplitude
estimation have been proposed [17,34,28,1,42]. All of those variants rely on the fact that
Qk A |0⟩n |0⟩ = cos((2k + 1)θa ) |ψ0 ⟩n |0⟩ + sin((2k + 1)θa ) |ψ1 ⟩n |1⟩ , (36)
where θa is an angle that satisfies the equation a = sin2 (θa ). The main difference between the algo-
rithms is the way they select the specific powers of k and the way they combine the measurements
to arrive at the result. In our work, we focused on the three algorithms implemented in Qiskit:
Iterative Amplitude Estimation (IAE) [17], Maximum Likelihood Amplitude Estimation (MLAE)
[34], and Faster Amplitude Estimation (FAE) [28]. After some primary experiments, we determined
that IAE appeared the most promising.
The IAE algorithm searches for θa within the interval [θl , θu ] ⊆ [0, π2 ]. IAE relies on the fact
that the probability of measuring |1⟩ in the result qubit is equal to
P[|1⟩] = sin2 ((2k + 1)θa ). (37)

Electronic copy available at: https://ssrn.com/abstract=4137397

16 R. Pracht et al.

After exploiting some trigonometric identities, this probability can be transformed to

P[|1⟩] = cos((4k + 2)θa ). (38)

Since cosine cannot be inverted without the knowledge if the angle belongs to [0, π] or [π, 2π], the
algorithm takes a different approach. Its actual goal is to find the largest value of k for which the
scaled interval [(4k + 2)θl , (4k + 2)θu ]mod2π is fully contained in either [0, π] or [π, 2π]. With this
knowledge, cos((4k + 2)θa ) can be inverted, providing a good estimate for θa and therefore a good
estimate for a.
The Qiskit implementation of IAE relies on two parameters: α ∈ (0, 1), which determines the
confidence interval 1 − α, and ϵ > 0, which denotes the target accuracy. These two values are used
to determine the maximum number of shots Nmax and influence the overall number of iterations.
In our case, the values for these two parameters were selected experimentally (ϵ = 0.001, α = 0.05).

3.6 Post-processing
The final stage is quite easy, we only have to convert the probability of the state |1⟩ into the price
of the Asian option:
p− 1
fmax c 2 + 1
price = . (39)
2

4 Result evaluation
In this section we examine the performance of Fx Asian call Option algorithm evaluated on quantum
simulator. The EUR/PLN exchange rate was modeled using the same stochastic differential equation
8 but we choose five different strikes from deep "out of the money" to "in the money". In that case,
we can evaluate how our algorithm performs in different market situations. During our work, we’ve
seen that there are three main sources of error:
– The discretization error introduced by the procedure described in a 3.1.
– The pay-off calculation error, whose main source is the sin2 approximation, see 3.4.
– Quantum Amplitude Estimation (QAE), see 3.5.
Therefore, we analyzed the impact of each kind of error on the final price in the different strikes.

4.1 Problem instance

We have broken the whole algorithm into three steps which make it easier to see where the error is
introduced.
1. We used the algorithm described in 3.1 to generate the EUR/PLN exchange rates paths, next we
calculate the option price classically. So in that stage, only the discretization error is included.
2. The second stage is to calculate the option price by the algorithm described in 3.1, 3.2, 3.3, 3.4
and 3.6. We don’t use QAE to extract the probability of state |1⟩ but we use a state vector
simulator instead. So here, we can perfectly estimate the |1⟩ probability, and therefore the price
of the evaluated Asian option includes only the discretization and sinus approximation errors.
3. The whole algorithm using also Quantum Amplitude Estimation (QAE), see 3.5.

Electronic copy available at: https://ssrn.com/abstract=4137397

 Fx Asian Option Pricing using Quantum Computers 17

4.2 Comparison with classical results

The results of evaluating the option pricing algorithm for different strike values can be found in
Table 1.

Table 1: The results of the option pricing for different strike values.
Strike Classical MC Discretization State vector QAE
4.1639 554 135.97 553 930.15 554 053.05 589 825.94
4.3952 333 190.20 332 950.70 333 259.19 364 011.45
4.6265 140 896.63 140 640.47 141 219.63 142 442.56
4.8578 32 870.59 31 701.01 32 424.13 94 697.46
5.0892 3 579.75 3 431.16 4 061.46 27 998.4

We can see that except for the strike value 5.0892, all of the other evaluations are similar between
the classical Monte-Carlo and the results from the state vector. To tune the results we can choose
one of the two options:
1. Increase the number of qubits N used to store the underlying asset probability distribution
modeled by a stochastic differential equation, see Equation 13.
2. Decrease the c parameter, so our sin2 approximation is better, see Equation 31. The drawback
of this is that the small change in evaluating the probability of 1 has a big impact on the final
option value, so this is much more challenging for QAE.
The results of evaluating the option pricing algorithm for different parameters c and N for the
strike 5.0892 can be found in Table 2.

Table 2: Impact of algorithm parameters on the result for option with strike equal to 5.0892.
c N Classical MC Discretization State vector
0.05 3 3 579.75 3 431.16 4 061.46
0.05 4 3 579.75 3 228.29 3 778.65
0.01 3 3 579.75 3 431.16 3 456.39
0.01 4 3 579.75 3 228.29 3 250.31

Decreasing the c parameter during the Quantum Monte Carlo algorithm decreases the difference
between the option value obtained by state vector and classic algorithm using the same discretiza-
tion. This is as expected, the question is why increasing the number of qubits N in the distribution
loading algorithm increases the error. This should have the opposite impact. However, it should be
noted that this effect does occur for the option deep out of the money which are usually considered
as the hardest to price. Nevertheless, this requires further research.
Extracting the probability of |1⟩ state in the pay-off qubit using QAE is more challenging. As
we can see in table 3 only the valuation for strikes 4.1639, 4.3952 and 4.6265 are similar to classical
Monte Carlo. What is worse, not only the final estimate is far from expectations but also the
difference between the minimum and maximum values in algorithm executions is much larger. So

Electronic copy available at: https://ssrn.com/abstract=4137397

18 R. Pracht et al.

even if the algorithm will be executed multiple times is hard to estimate the correct value of the
price, please see table 3.

Table 3: The results of the option pricing for different strike values for QAE.
Strike Classical MC Min Avg Max
4.1639 554 135.97 559 535.99 589 825.94 627 083.38
4.3952 333 190.20 321 729.64 364 011.45 430 522.03
4.6265 140 896.63 141 521.11 142 442.56 143 943.22
4.8578 32 870.59 44 926.82 94 697.46 163 125.44
5.0892 3 579.75 15 685.7 27 998.4 45 607.69

5 Discussion and Conclusion

In this work, we have presented a quantum algorithm for pricing the currency Asian style option.
We have assumed that the distribution of the EUR/PLN currency under the martingale measure
is described by the Black-Scholes-Merton model. Contrary to previous work we don’t assume that
interest rate and volatility are constants and our algorithm allows as to be easily extended to the
local volatility model. This is a very important fact from the practical point of view because the
volatility smile is usually observed on the financial markets and the local volatility model is one
of the most important in the practical pricing of financial derivatives. The quadratic speed-up
of quantum Monte-Carlo was shown in the previous papers [29,45,33], we focus on the practical
aspect of the financial derivative evaluation. We compared the result of the Quantum Monte-Carlo
algorithm with the classical one we show that the results are similar for all Strike values.
According to our knowledge, this is the first paper to pricing Asian style options using four-time
points and without using the ancilla qubits owing to using the Quantum Fourier transform adder.
Is worth noting that in the practical application we usually need one point per trading day so using
about 300 points is nothing uncommon. Therefore algorithm should effectively do the summation,
the use of the Quantum Fourier transform for addition meets this challenge.
Overall our results look quite accurate compared to the classical one for all strikes except the
option deep out of the money. So we will continue our research to find out why that happen and
how we could improve that.
We also encountered more troubles with QAE algorithms, and the results are quite unstable. So
we will continue to work on it.

6 Acknowledgments
We acknowledge Professor Dariusz Gątarek for paper review.

References
1. Aaronson, S., Rall, P.: Quantum Approximate Counting, Simplified. Symposium on Simplicity in Al-
gorithms pp. 24–32 (Jan 2020). https://doi.org/10.1137/1.9781611976014.5

Electronic copy available at: https://ssrn.com/abstract=4137397

 Fx Asian Option Pricing using Quantum Computers 19

2. ANIS, M.S., Abraham, H., AduOffei, Agarwal, R., et al.: Qiskit: An open-source framework for quantum
computing (2021). https://doi.org/10.5281/zenodo.2573505
3. Beckman, D., Chari, A.N., Devabhaktuni, S., Preskill, J.: Efficient networks for quantum factoring.
Physical Review A 54(2), 1034–1063 (Aug 1996). https://doi.org/10.1103/physreva.54.1034
4. Black, F., Scholes, M.: The Pricing of Options and Corporate Liabilities. Journal of Political Economy
81(3), 637–654 (1973). https://doi.org/10.1086/260062
5. Boyle, P.P.: Options: A Monte Carlo approach. Journal of Financial Economics 4(3), 323–338 (1977).
https://doi.org/10.1016/0304-405X(77)90005-8
6. Brassard, G., Høyer, P., Mosca, M., Tapp, A.: Quantum amplitude amplification and esti-
mation. In: Lomonaco, Jr., S.J., Brandt, H.E. (eds.) Quantum Computation and Informa-
tion, Contemporary Mathematics, vol. 305, pp. 53–74. American Mathematical Society (2002).
https://doi.org/10.1090/conm/305/05215
7. Castagna, A.: FX Options and Smile Risk. The Wiley Finance Series, John Wiley & Sons (Feb 2010)
8. Cox, J.C., Ross, S.A., Rubinstein, M.: Option pricing: A simplified approach. Journal of Financial
Economics 7(3), 229–263 (1979). https://doi.org/10.1016/0304-405X(79)90015-1
9. Cuccaro, S.A., Draper, T.G., Kutin, S.A., Moulton, D.P.: A new quantum ripple-carry addition circuit
(2004), https://arxiv.org/abs/quant-ph/0410184
10. Curran, M.: Valuing Asian and Portfolio Options by Conditioning on the Geometric Mean Price. Man-
agement Science 40(12), 1705–1711 (1994), http://www.jstor.org/stable/2632947
11. Draper, T.G.: Addition on a Quantum Computer (2000), https://arxiv.org/abs/quant-ph/0008033
12. Draper, T.G., Kutin, S.A., Rains, E.M., Svore, K.M.: A Logarithmic-Depth Quantum Carry-Lookahead
Adder. Quantum Info. Comput. 6(4), 351–369 (jul 2006). https://doi.org/10.26421/QIC6.4-5-4
13. Dupire, B.: Pricing with a Smile. Risk Magazine 7(1), 18–20 (1994)
14. Forsyth, P., Vetzal, K., Zvan, R.: Convergence of numerical methods for valuing path-
dependent options using interpolation. Review of Derivatives Research 5, 273–314 (02 2002).
https://doi.org/10.1023/A:1020823700228
15. Gossett, P.: Quantum Carry-Save Arithmetic (1998), https://arxiv.org/abs/quant-ph/9808061
16. de la Grand’rive, P.D., Hullo, J.F.: Knapsack Problem variants of QAOA for battery revenue optimi-
sation (2019), https://arxiv.org/abs/1908.02210
17. Grinko, D., Gacon, J., Zoufal, C., Woerner, S.: Iterative quantum amplitude estimation. npj Quantum
Information 7, 52 (03 2021). https://doi.org/10.1038/s41534-021-00379-1
18. Grover, L., Rudolph, T.: Creating superpositions that correspond to efficiently integrable probability
distributions (2002). https://doi.org/10.48550/ARXIV.QUANT-PH/0208112
19. Hull, J.C., White, A.D.: Efficient procedures for valuing european and american path-dependent options.
The Journal of Derivatives 1(1), 21–31 (1993). https://doi.org/10.3905/jod.1993.407869
20. Ingersoll, J.E.: Theory of Financial Decision Making. Rowman & Littlefield (1987)
21. Iten, R., Colbeck, R., Kukuljan, I., Home, J., Christandl, M.: Quantum circuits for isometries. Phys.
Rev. A 93, 032318 (Mar 2016). https://doi.org/10.1103/PhysRevA.93.032318
22. Kemna, A., Vorst, A.: A pricing method for options based on average asset values. Journal of Banking &
Finance 14(1), 113–129 (1990). https://doi.org/https://doi.org/10.1016/0378-4266(90)90039-5, https:
//www.sciencedirect.com/science/article/pii/0378426690900395
23. Kitaev, A.Y.: Quantum measurements and the Abelian Stabilizer Problem (1995), https://arxiv.
org/abs/quant-ph/9511026
24. Levy, E.: Pricing European Average Rate Currency Options. Journal of International Money and Fi-
nance 11(5), 474–491 (1992). https://doi.org/10.1016/0261-5606(92)90013-N
25. Merton, R.C.: Theory of Rational Option Pricing. The Bell Journal of Economics and Management
Science 4(1), 141–183 (1973). https://doi.org/10.2307/3003143
26. Meter, R.V., Munro, W.J., Nemoto, K., Itoh, K.M.: Arithmetic on a Distributed-Memory
Quantum Multicomputer. J. Emerg. Technol. Comput. Syst. 3(4), 1–23 (jan 2008).
https://doi.org/10.1145/1324177.1324179

Electronic copy available at: https://ssrn.com/abstract=4137397

20 R. Pracht et al.

27. Milevsky, M.A., Posner, S.E.: Asian options, the sum of lognormals, and the reciprocal gamma distri-
bution. The Journal of Financial and Quantitative Analysis 33(3), 409–422 (1998), http://www.jstor.
org/stable/2331102
28. Nakaji, K.: Faster Amplitude Estimation. Quantum Information and Computation 20(13 & 14), 1109–
1123 (Nov 2020). https://doi.org/10.26421/qic20.13-14-2
29. Rebentrost, P., Gupt, B., Bromley, T.R.: Quantum computational finance: Monte carlo pricing of finan-
cial derivatives. Phys. Rev. A 98, 022321 (Aug 2018). https://doi.org/10.1103/PhysRevA.98.022321
30. Rogers, L.C.G., Shi, Z.: The Value of an Asian Option. Journal of Applied Probability 32(4), 1077–1088
(1995), http://www.jstor.org/stable/3215221
31. Ruiz-Perez, L., Garcia-Escartin, J.C.: Quantum arithmetic with the quantum Fourier transform. Quan-
tum Information Processing 16(6) (Apr 2017). https://doi.org/10.1007/s11128-017-1603-1
32. Shreve, S.: Stochastic Calculus for Finance II: Continuous-Time Models. Springer Finance, Springer
New York (2004)
33. Stamatopoulos, N., Egger, D.J., Sun, Y., Zoufal, C., Iten, R., Shen, N., Woerner, S.: Option Pricing
using Quantum Computers. Quantum 4, 291 (Jul 2020). https://doi.org/10.22331/q-2020-07-06-291
34. Suzuki, Y., Uno, S., Raymond, R., Tanaka, T., Onodera, T., Yamamoto, N.: Amplitude
estimation without phase estimation. Quantum Information Processing 19(2) (Jan 2020).
https://doi.org/10.1007/s11128-019-2565-2
35. Takahashi, Y., Kunihiro, N.: A Fast Quantum Circuit for Addition with Few Qubits. Quantum Infor-
mation and Computation 8(6 & 7), 636–649 (jul 2008). https://doi.org/10.26421/QIC8.6-7-5
36. Takahashi, Y., Tani, S., Kunihiro, N.: Quantum Addition Circuits and Unbounded Fan-Out. Quantum
Information and Computation 10(9 & 10), 872–890 (sep 2010). https://doi.org/10.26421/QIC10.9-10-
12
37. Trisetyarso, A., Meter, R.V.: Circuit Design for A Measurement-Based Quantum Carry-
Lookahead Adder. International Journal of Quantum Information 08(05), 843–867 (Aug 2010).
https://doi.org/10.1142/s0219749910006496
38. Turnbull, S.M., Wakeman, L.M.: A Quick Algorithm for Pricing European Average Options. Journal
of Financial and Quantitative Analysis 26(3), 377–389 (1991). https://doi.org/10.2307/2331213
39. Vedral, V., Barenco, A., Ekert, A.: Quantum networks for elementary arithmetic operations. Physical
Review A 54(1), 147–153 (Jul 1996). https://doi.org/10.1103/physreva.54.147
40. Večeř, J.: A new pde approach for pricing arithmetic average asian options. J. Comput. Finance 4 (01
2001). https://doi.org/10.21314/JCF.2001.064
41. Vorst, T.: Prices and hedge ratios of average exchange rate options. International Review of Financial
Analysis 1(3), 179–193 (1992). https://doi.org/10.1016/1057-5219(92)90003-M
42. Wie, C.R.: Simpler Quantum Counting. Quantum Information and Computation 19(11 & 12), 967–983
(Sep 2019). https://doi.org/10.26421/QIC19.11-12-5
43. Wiebe, N., Roetteler, M.: Quantum Arithmetic and Numerical Analysis Using Repeat-until-
Success Circuits. Quantum Information and Computation 16(1 & 2), 134–178 (jan 2016).
https://doi.org/10.26421/QIC16.1-2-9
44. Wilmot, P., Dewynne, J., Howison, S.: Option Pricing: Mathematical Models and Computation. Oxford
Financial Press (1993)
45. Woerner, S., Egger, D.J.: Quantum risk analysis. npj Quantum Information 5(1) (Feb 2019).
https://doi.org/10.1038/s41534-019-0130-6
46. Wystup, U.: FX Options and Structured Products. The Wiley Finance Series, John Wiley & Sons
(2006)
47. Zoufal, C., Lucchi, A., Woerner, S.: Quantum generative adversarial networks for learning and loading
random distributions. npj Quantum Information 5(1) (nov 2019). https://doi.org/10.1038/s41534-019-
0223-2

Electronic copy available at: https://ssrn.com/abstract=4137397