applied

sciences
Article
Beyond Stochastic Gradient Descent for Matrix
Completion Based Indoor Localization
Wafa Njima 1,2, *, Rafik Zayani 1,2 , Iness Ahriz 2 , Michel Terre 2 and Ridha Bouallegue 1
1 University of Carthage, Higher School of Communication of Tunis, LR-11/TIC-03 Innov’COM Laboratory,
2083 Ariana, Tunisia; [email protected] (R.Z.); [email protected] (R.B.)
2 Conservatoire National des Arts et Métiers, CEDRIC/ LAETITIA Laboratory, 75003 Paris, France;
[email protected] (I.A.); [email protected] (M.T.)
* Correspondence: [email protected]




Received: 7 April 2019; Accepted: 8 June 2019; Published: 13 June 2019


Abstract: In this paper, we propose a high accuracy fingerprint-based localization scheme for the
Internet of Things (IoT). The proposed scheme employs mathematical concepts based on sparse
representation and matrix completion theories. Specifically, the proposed indoor localization scheme
is formulated as a simple optimization problem which enables efficient and reliable algorithm
implementations. Many approaches, like Nesterov accelerated gradient (Nesterov), Adaptative
Moment Estimation (Adam), Adadelta, Root Mean Square Propagation (RMSProp) and Adaptative
gradient (Adagrad), have been implemented and compared in terms of localization accuracy and
complexity. Simulation results demonstrate that Adam outperforms all other algorithms in terms of
localization accuracy and computational complexity.

Keywords: Adadelta; adaptative gradient; adaptative moment estimation; Gradient descent; indoor
localization; matrix completion; Nesterov accelerated gradient; root mean square propagation;
trilateration

1. Introduction
The concept of the Internet of Things (IoT), where objects, with their own identifiers, have the
ability to transfer data over a network without requiring human interaction, is federating more and
more interests nowadays [1,2]. In addition to communication technology, data management and data
privacy and security, the development of smart applications is strongly related to the notion of physical
location and positions [3–5]. Therefore, the infrastructure has to support finding things according to
location taking mobility into account. Localization technologies will then play a crucial role for the
future of IoT and may be directly embedded into the infrastructure or into “things”.
The Global Positioning System (GPS) is the most popular localization solution allowing
positioning, with a high accuracy, when at least four satellites are available [6]. In addition, a high signal
quality from those satellites is necessary to perform localization and this is why such a solution cannot
easily be deployed in an indoor environment. An alternative solution is to use the communication
infrastructure to perform localization. Different methods have been proposed and studied. The radio
fingerprinting is a well known localization technique which is organized in two steps [7,8]. In the
first step, a radio map is constructed and, in the second step, online measurement is compared to
the radio map [9]. The communication signal can also be used to perform the trilateration [10–12]
where the distances between the target and reference positions can be estimated by using the Received
Signal Strength Indicator (RSSI) [13]. The performance of such a localization technique depends on the
number of detected reference positions.
In indoor environments, when the propagation conditions are very severe, the RSSI is often too
weak to be correctly detected by the target receiver. So, only partial pairwise inter sensor nodes

Appl. Sci. 2019, 9, 2414; doi:10.3390/app9122414 www.mdpi.com/journal/applsci

Appl. Sci. 2019, 9, 2414 2 of 21

distances can be calculated from the measured RSSI values. To overcome this problem, we propose in
this paper, to estimate the complete distance matrix using matrix completion algorithm. This approach
aims to approximate the distance between the target and all the reference positions through the spatial
correlation structure of the fingerprints. After the matrix completion, the localization can be performed
using either matrix decomposition or fingerprinting [14–17]. In this paper, given a set of reference
nodes and the RSSI information between sensors, we process sensor localization using the trilateration
technique. To the best of our knowledge, it is the first time that matrix completion algorithms are
combined with trilateration. So, we propose to improve the localization accuracy of the trilateration
technique using a complete pairwise squared Euclidean distance matrix instead of only using a partial
number of pairwise distances calculated from measured RSSI only.

1.1. Related Works

As mentioned in the previous section, it is difficult, from available RSSI measurements, to acquire
all the pairwise distances between sensors and to obtain a full connected squared distance matrix
so called the Euclidean distance matrix (EDM) [18]. This is due to the limitation of communication
range and multipath effects. To provide a reliable distance information to each sensor node, the matrix
completion is then proposed and its role is to recover the complete Euclidean distance matrix EDM
from an incomplete matrix.
Several algorithms and approaches have been proposed and studied to approximate the missing
distances for RSSI measurements based systems. Authors in [14] formulate the matrix completion
problem using the least squared minimization. To solve the minimization problem, they introduce
a modified iterative Newton’s algorithm [19] to optimize the objective function. This solution can
be very fast if the parameters are well chosen. The main drawback is that the result is very sensitive
to initialization. It can be considered as a good initial location estimation for other fine localization
algorithms. Other existing approaches based on Singular Value Thresholding (SVT) [20], Accelerated
Proximal Gradient (APG) [21] and Augmented Lagrange Multiplier (ALM) [22] have been proposed
in recent years. SVT method which relies on the Singular Value Decomposition (SVD) of a low
rank matrix is associated to poor completion performance and slow convergence rate. Motivated
by the above reasons, ALM and APG methods have been proposed. However, these methods
depend on the choice of relevant parameters and are sensitive to noise. To overcome this problem,
a regularized matrix completion model is proposed in [23] introducing the multivariate function
Bregman divergence to solve the EDM problem. The major drawback of such a centralized method is
its computational cost. The aforementioned methods are all based on RSSI. As an alternative to RSSI
measurements, Nguyen et al. focus on connectivity/distance information setting a routing protocol
in place [15,16]. They formulate the matrix completion problem as an unconstrained optimization in
Smooth Riemannian manifold. Then, the nonlinear conjugate gradient algorithm is applied on the
smooth Riemannian manifold. This approach recovers the EDM in noiseless and noisy environments.
However, it focused only on two dimensional position information. Authors in [24] cover the unknown
measurements under the existence of noise which is classified into three categories. This approach
achieves a good localization accuracy compared to other existing ones. However, it is a centralized
localization approach. Motivated by the above reasons, based on RSSI measurements, we solve the
minimization problem using advanced algorithms to reduce the running time and the computational
complexity while ensuring a good localization accuracy.
After completing the EDM, the second step is to estimate the location of sensors. For this,
the completed pairwise distance information can be transformed into the estimated coordinates of
sensor nodes by applying a factorization process [14]. This factorization requires that the matrix is
Semi Definite Positive (SDP). The SDP problem can be efficiently and accurately solved by CVX
toolbox on Matlab [25]. However, if the matrix is not SDP, the problem can also be solved by
introducing Semidefinite Relaxations (SDR) [26,27]. The SDR based localization scheme in [26] relaxes
the non convex localization into a convex one and marks a good localization accuracy. However,
Appl. Sci. 2019, 9, 2414 3 of 21

the computational complexity of SDR-based techniques is closely related to the problem size. Therefore,
this algorithm is adapted for medium size network only due to the highly required running time.
To reduce the running time, a weighted semidefinite relaxation localization method is carried out
in [27]. It aims to improve the accuracy of localization, but it is still not suitable for dealing with large
networks. A multiple sources localization problem is formulated in [28] as a unimodality constrained
matrix factorization (UMF) and two rotation techniques have been developed to solve the problem.
In [29], an eigendecomposition is first applied to find the local locations of sensor nodes, then a
rotation matrix and a translation vector are used to transform them to true locations. Zhang et al.
in [30] proposed to estimate the positions of nodes by the classic multidimensional scaling algorithm
(MDS) [31] using a truncated eigendecomposition. However, the MDS method requires high recovery
rate of the EDM which is not guaranteed when working in noisy environments. The localization can
be also ensured by the fingerprinting technique. The principal drawback is that an offline radio map
construction is needed. Furthermore, due to environment changes this radio map has to be frequently
updated. In this paper, we propose to use the trilateration method. It can be applied if we have at least
three pairwise known distances between the sensors if the localization is performed in 2D space. At
each sensor node to be localized, data fusion is conducted by combining measurements from different
nodes to estimate its location.

1.2. Contribution
The contribution of our work is threefold:

• Improving the localization accuracy of trilateration technique: We develop high accuracy

fingerprint based indoor localization scheme which is based on sparse representation and matrix
completion theories. As said before, trilateration technique is based on pairwise distances
between the target node and the anchor nodes. Due to the fact that propagation conditions
are not optimal in an indoor environment, several pairwise distances cannot be measured. So,
our main contribution is to enhance the localization’s accuracy of trilateration by estimating all
pairwise distances.
• Matrix completion resolution: We formulate the indoor localization scheme as a simple
optimization problem which enables efficient and reliable algorithm implementation.
• Solving the proposed optimization problem: We develop closed-form algorithms, which can be
reproduced by simple implementation, to solve the indoor localization problem. Specifically,
we adopt recent methods like Nesterov accelerated gradient (Nesterov), adaptative Moment
Estimation (Adam), Adadelta, Root Mean Square Propagation (RMSProp) and Adaptative
gradient (Adagrad).

The remainder of this paper is organized as follows: In Section 2, we present the system model.
In Section 3, details of our contribution and different algorithms are given. Obtained simulation results
are presented and discussed in Section 4. The method verification using real measurements is detailed
in Section 5. Finally, we conclude the paper in Section 6.
Notations: The following notations are used throughout our paper. (.) T is the transpose operation.
||.||2 , ||.|| F and ||.||∗ denote the l2 norm, the Frobenius norm and the nuclear norm, respectively.
diag(A) returns a column vector of the diagonal elements of the matrix A. ∇( f (.)) is the gradient of
function f . A(t) is the matrix obtained at iteration t. A · B is the product of matrices. A  B is the
Hadamard product of A and B. Matrices are presented by non italic bold capital letters. Furthermore,
variables are presented by italic lowercase characters.

2. System Model
In a system where our approach is adopted, each sensor node uses the following 2 steps to find
its position (Figure 1).

• Step 1: Refine and complete the squared distance matrix.

Appl. Sci. 2019, 9, 2414 4 of 21

• Step 2: Once the matrix is completed, the coordinates of the node can be retrieved using the classic
trilateration process.

RSSI measurements corresponding to different sensor nodes are used to construct an RSSI matrix.
From the RSSI values, the Euclidean Distance Matrix containing the distance information between
each pair of sensor nodes X is built. Due to the limitation of radio communication range, the matrix of
RSSI measurements is partially known. Thus, the matrix X is incomplete (only a small number of X
entries are available) and can be affected by noise. This incomplete matrix can not efficiently serve
for localization. It should be completed. Let us define the matrix Xtrue which is the complete EDM.
Therefore, the problem is how to recover the unknown elements of X giving a small number of known
entries of X.

RSSI matrix (7x7)

A user to locate (unknown node Ui)
R_U1 R_U1 R_U1 R_U1
? ? ?
_U1 _U2 _A1 _A2

R_U2 R_U2 R_U2 R_U2

? ? ?
_U1 _U2 _U3 _A1
A reference position (anchor node Ai)
R_U3 R_U3 R_U3 R_U3
? ? ?
_U2 _U3 _A2 _A3
RSSI values received from
Not detected R_U4 R_U4 R_U4
Ui at U1 ? ? ?
_U4
?
_A2 _A3
node From RSSI to
U4 U1 R_A1 R_A1 R_A1 distance using a
? ? ? ?
0
R_U1 R_U1 R_U1 _U1 _U2 _A1 propagation
_U2 ? ? _A1 _A2 ?
model
R_A2 R_A2 R_A2 R_A2 R_A2
? ?
A2 _U1 _U3 _U4 _A2 _A3
A1 RSSI values
received from Ai ? ?
R_A3 R_A3
?
R_A3 R_A3
at U1 _U3 _U4 _A2 _A3

U2
A3 U3

(d_U1 (d_U1 (d_U1 (d_U1 (d_U1 (d_U1 (d_U1 (d_U1 (d_U1

0 0 ? ? ?
_U2)² _U3)² _U4)² _A1)² _A2)² _A3)² _U2)² _A1)² _A2)²

(d_U2 (d_U2 (d_U2 (d_U2 (d_U2 (d_U2 (d_U2 (d_U2 (d_U2

0 0 ? ? ?
_U1)² _U3)² _U4)² _A1)² _A2)² _A3)² _U1)² _U3)² _A1)²
Step 2 Step 1
(d_U3 (d_U3 (d_U3 (d_U3 (d_U3 (d_U3 (d_U3 (d_U3 (d_U3
0 ? 0 ? ?
_U1)² _U2)² _U4)² _A1)² _A2)² _A3)² _U2)² _A2)² _A3)²
Matrix
Trilateration (d_U4 (d_U4 (d_U4 (d_U4 (d_U4 (d_U4 (d_U4 (d_U4
_U1)² _U2)² _U3)²
0
_A1)² _A2)² _A3)² completion ? ? ? 0 ?
_A2)² _A3)²
process
process
(d_A1 (d_A1 (d_A1 (d_A1 (d_A1 (d_A1 (d_A1 (d_A1
0 ? ? 0 ? ?
_U1)² _U2)² _U3)² _U4)² _A2)² _A3)² _U1)² _U2)²

(d_A2 (d_A2 (d_A2 (d_A2 (d_A2 (d_A2 (d_A2 (d_A2 (d_A2 (d_A2
0 ? ? 0
_U1)² _U2)² _U3)² _U4)² _A1)² _A3)² _U1)² _U3)² _U4)² _A3)²

(d_A3 (d_A3 (d_A3 (d_A3 (d_A3 (d_A3 (d_A3 (d_A3 (d_A3

0 ? ? ? 0
_U1)² _U2)² _U3)² _U4)² _A1)² _A2)² _U3)² _U4)² _A2)²
User localized (x,y)
Estimated matrix (7x7) Observed matrix (7x7)
Reconstructed EDM Incomplete and noisy EDM

Figure 1. Algorithm details considering the sensor network consisting of seven sensor nodes from
which three are anchors.

Assume that m is the number of nodes with known positions named ’Anchor nodes’. In contrast,
(n − m) is the number of sensors with unknown positions. n is the total number of sensor nodes
(anchor nodes and unknown nodes) placed in the indoor environment. Ui is the ith unknown node
where i = 1, 2, ..., (n − m). A j is the jth anchor node where j = 1, 2, ..., m. X is the (n × n) Euclidean
distance matrix. It can be partitioned as follows:
" #
X11 X12
X= , (1)
X21 X22

where X22 is the distance sub matrix between each pair of anchor nodes. It is obtained by calculating the
exact distance between each pair of anchors. d Ai A j = ||C Ai − C A j ||2 is the pairwise distance between
anchor node i (Ai ) and anchor node j (A j ). C Ai ∈ R3 are the location coordinates of anchor node i.
X11 is the (n − m) × (n − m) distance sub matrix between each pair of unknown nodes. X12 and X21 ,
where X12 = X21 T , are the distance sub matrices between each pair of anchors and unknown nodes.
Appl. Sci. 2019, 9, 2414 5 of 21

X11 , X12 and X21 are obtained from RSSI measurements using the log normal shadowing propagation
model which is used to express the pathloss measurements in dBs [32]:

d
plij = pl0 + 20log10 ( f ) + 10$log10 ( ), (2)
d0

where pl0 is the pathloss value at a reference distance d0 , $ is the pathloss exponent, f is the used
frequency, d is the distance between node i and node j. The used propagation parameters are
defined later in Section 4. Since pairwise distances are dependant because of the dependency of
RSSI in the indoor environments, this matrix should have low rank which motivates the use of matrix
completion algorithms.
After completing the Euclidean distance matrix, a trilateration process is adopted by each
unknown node in order to estimate its location [10]. A reminder of this method is introduced below.
A combination of estimated distances is required. In this combination process, we use only the distance
estimated submatrix X̂12 which contains the distances between the unknown nodes and all anchor
nodes. The completion of the total matrix brings more distance information. Distances are estimated
from each other due to the fact that columns are dependant (low rank matrix).

d211 d212 d21m

 
. .

 d221 d222 . . d22m 
X̂12 = . . . . . , (3)
 
. . . . .
 
 
d2(n−m)1 d2(n−m)2 . . d2(n−m)m

Let (xˆj , yˆj ) be the estimated coordinates of an unknown node; Ĉj = [xˆj , yˆj ] T . (xi , yi ), (i = 1, 2, ..., m)
are the coordinates of the anchor nodes. d2ij = (xi − x̂ j )2 + (yi − ŷ j )2 where j = (1, 2, ..., n − m) and
k2i = x2i + y2i . The estimated coordinates are calculated by the following equation

Ĉ = (B T B)−1 B T P, (4)

k22 − k21 − d22j + d21j

   
x2 − x1 y2 − y1
 k2 − k2 − d2 + d2  x − x y3 − y1 
 3 1 3j 1j   3 1 
where P = 21  ......  And B =  . .
   

   
 ......   . . 
2 2 2 2
km − k1 − dmj + d1j xm − x1 y m − y1

3. Proposed Matrix Completion Based Localization

In this section, we formulate our proposed approach as a convex optimization problem which is
resolved using the developed closed form algorithms via Gradient descent and its variants.

3.1. Problem Formulation

Our goal is to reconstruct the complete distance matrix from incomplete and noisy data. The problem
of recovering a low rank matrix from a small number of known entries is known as:
(
min rank (X̂)
X̂ , (5)
s.t. x̂ij = xij where i, j ∈ ω
ω is the set of known entries. Due to the non convexity and non linearity of the rank matrix [33],
the problem in Equation (1) cannot be solved numerically. Inspired by the theory of Compressed
Sensing (CS), Candes and Recht proposed to replace the rank function in Equation (1) by the nuclear
norm [34]. The model in (1) is reformulated into:
Appl. Sci. 2019, 9, 2414 6 of 21

(
min ||X̂||∗
X̂ , (6)
s.t. x̂ij = xij where i, j ∈ ω

||X̂||∗ is the sum of the singular values of X̂ (i.e., ||X̂||∗ = ∑nj=1 s j . Where X̂ = USV T ). Considering the
assumption of low rank (r << n) and taking into account that observations are usually affected by
noise, the model of matrix completion can be defined as:

min λ × ||X̂||∗ + ||H  (X̂ − X)||2F , (7)

X̂
q
λ is a tunable parameter and ||.|| F is the Frobenius norm. ||X̂|| F = ∑in=1 ∑nj=1 | x̂ij |2 . H is the matrix
whose entries are:
(
1 if (i, j) ∈ ω
hij = h ji = , (8)
0 otherwise
We denote the objective function as:

J (X̂) = f (X̂) + λ × l (X̂), (9)

where f (X̂) = ||H  (X̂ − X)||2F and l (X̂) = ||(1 − H)  X̂||∗ . The defined optimization problem can
be solved efficiently by using iterative Gradient Descent method and its variants. The developed
algorithm is summarized in Table 1. Where V(t) is the matrix used to update the distance matrix, index
t refers to the number of update iteration. Many approaches, detailed in the next section, have been
adopted to find the matrix update V(t) .

Table 1. The problem described in four steps.

1. Input: X, H, number max of iterations

2. Initialization: X̂(0) , t
3. while t <number max of iterations
Update X̂:

X̂(t+1) = X̂(t) − V(t) ,

t = t+1
end
4. Return X̂

3.2. Matrix Completion: Optimization over GD and Its Variants

In the following, we review the different optimization methods that are widely used by the deep
learning community to update the localization matrix X̂: Gradient descent (GD), Nesterov accelerated
gradient (NAG), Adaptive Gradient (Adagrad), Root Mean Square Propagation (RMSProp), Adadelta
and Adaptive Moment Estimation (Adam). We discard the class of algorithms that are computationally
very expensive for high dimensional data sets, e.g., the second-order Newton’s method [35].

3.2.1. Gradient Descent (GD)

Gradient descent is an iterative method that aims to find local minimum of differentiable cost
functions [36]. It is the most common first-order optimization algorithm in machine learning and deep
learning. GD is based on updating each element of matrix X̂(t) in the direction to optimize the objective
function J (X̂(t) ). The new parameter V(t) can be adjusted as

V(t) = α∇( J (X̂(t) )), (10)

Appl. Sci. 2019, 9, 2414 7 of 21

α is the learning rate from range (0, 1). ∇( J (X̂(t) )) is the gradient of the cost function with respect to
the parameter matrix. It can be computed as follows:

∇( J (X̂(t) )) = ∇( f (X̂(t) )) + λ × ∇(l (X̂(t) )), (11)

where

∇( f (X̂(t) )) = U(t) = ∇(||H  (X̂(t) − X(t) )||2F )

= 2 × H  (X̂(t) − X(t) ), (12)

∇(l (X̂(t) )) is calculated as follows:

∇(l (X̂(t) )) = W(t) = ∇(||(1 − H)  X̂(t) ||∗ )

= (1 − H)  (X̂(t) · ((X̂(t) )T · X̂(t) )−0.5 ), (13)

We might need some sort of regularization because the inverse of the square root of (X̂(t) ) T · X̂(t)
may not exist, e.g.,

∇(l (X̂(t) )) = ∇(||(1 − H)  X̂(t) ||∗ ) = (1 − H)  (X̂(t) · ((X̂(t) )T · X̂(t) + e × I)−0.5 ), (14)

where e is a regularization parameter and I(n × n) is the identity matrix. Then

∇( J (X̂(t) )) = 2 × H  (X̂(t) − X(t) ) + λ × (1 − H)  (X̂(t) · ((X̂(t) )T · X̂(t) + e × I)−0.5 ), (15)

We obtain

V(t) = α × (2 × H  (X̂(t) − X(t) ) + λ × (1 − H)  (X̂(t) · ((X̂(t) ) T · X̂(t) + e × I)−0.5 )), (16)

Using the classic Gradient descent, the known entries are very well estimated, this is why it will be
used in the rest of the paper to estimate U(t) . To estimate W(t) , we propose to use the following algorithms.

3.2.2. Nesterov Accelerated Gradient (NAG)

A second commonly used variant is the acceleration of Nesterov, that has been published in [37]
and fits in the same vein as the idea of the Momentum [38]. It has the same intuition using the gradient
history but it calculates the gradient with respect to an approximate future values of our parameters
instead of the current parameters. To update W(t) with NAG, we use the following equation.

W(t) = µ × W(t−1) + λ × ∇(l (X̂(t) − µ × W(t−1) )), (17)

where

W(t−1) = ∇(l (X̂(t−1) )) = (1 − H)  (X̂(t−1) · ((X̂(t−1) ) T · X̂(t−1) + e × I)−0.5 ), (18)

And

∇(l (X̂(t) − µ × W(t−1) )) = ∇(l (X̂(t) − µ × (1 − H)  (X̂(t−1) · ((X̂(t−1) )T · X̂(t−1) + e × I)−0.5 ))), (19)
Appl. Sci. 2019, 9, 2414 8 of 21

Suppose that

Ŷ(t) = X̂(t) − µ × (1 − H)  (X̂(t−1) · ((X̂(t−1) ) T · X̂(t−1) + e × I)−0.5 ), (20)

Then,

∇(l (Ŷ(t) ) = (1 − H)  (Ŷ(t) · ((Ŷ(t−1) )T · Ŷ(t−1) + e × I)−0.5 ), (21)

We obtain

W(t) = µ × (1 − H)  (X̂(t−1) · ((X̂(t−1) ) T · X̂(t−1) + e × I)−0.5 )

+ λ × (1 − H)  (Ŷ(t) · ((Ŷ(t−1) )T · Ŷ(t−1) + e × I)−0.5 ), (22)

The main advantage of the NAG method compared to the GD one is related to the fact that an
anticipatory update prevents us from going too fast and results in increased responsiveness to the
landscape of the loss function [36].

3.2.3. Adaptive Gradient (Adagrad)

Duchi et al. [35] introduced Adagrad algorithm in the context of projected gradient method.
Adagrad aims to adapt the learning rate to the updated parameters, performing low learning rate
(i.e., smaller updates) when the memory of squared gradients is high and larger updates conversely.
Adagrad update rule is as follows:

• We set E(t) to be the gradient of the objective function with respect to the parameter X̂(t)

δl (X̂(t) )
E(t) =
δX̂(t)
= (1 − H)  (X̂(t) · ((X̂(t) )T · X̂(t) + e × I)−0.5 ), (23)

• we compute the memory of squared gradients over time as

t
G(t) = ∑ ( E (i ) )2 , (24)
i =1

• we modify the general learning rate α at each time step t for every parameter X̂(t) based on the
sum of the squares of the gradients that have been computed for X̂(t) up to time step t.

1
W(t) = p  E(t) , (25)
G(t) + e × I

while e is a regularizing term used to avoid division by zero. It is worth mentioning that we do
not need to adjust the learning rate.

3.2.4. Root Mean Square Propagation (RMSProp)

Tieleman et al. [39] introduced this algorithm in 2012. It is described in the following 2 steps.
Instead of using the memory of all squared gradients, RMSProp uses only recent past gradients
computed in a restricted time.

• We compute the local average of previous (E(t) )2 as

G̃(t) = ρ × G(t−1) + (1 − ρ) × (E(t) )2 , (26)

Appl. Sci. 2019, 9, 2414 9 of 21

• Then, we apply the update

1
W(t) = p  E(t) , (27)
G̃(t) + e × I

3.2.5. Adadelta
Adadelta was introduced in 2012 by Zeiler [39]. It aims at circumventing Adagrad’s weakness that
consists in its aggressive decreasing learning rate caused by accumulating all past squared gradients
in the denominator. Adadelta scales learning rate using only recent past gradients computed in a
restricted time (i.e., not the whole history). In addition, Adadelta uses an accelerator term by taking
into account past updates (as in Momentum). Adadelta update rule is as follows:

• we compute gradient E(t) as in Equation (23).

• we compute the local average G̃(t) of previous (E(t) )2
• we compute new term accumulating prior updates ( Momentum : acceleration term)

Z ( t ) = ρ × Z ( t −1) + (1 − ρ ) × ( W ( t −1) )2 , (28)

• Then, we apply the update
p
Z(t) + e × I
W(t) = p  E(t) , (29)
α G̃(t) + e × I

3.2.6. Adaptive Moment Estimation (Adam)

Another optimization method that computes adaptive learning rate for each parameter is
introduced by Kingma and Ba in [40]. Adam uses the first and the second moments of the gradients and
has strong similarities with Adadelta. Indeed, it uses the second gradient moment in the denominator
and a momentum term. Adam update rule consists of the following steps.

• Compute second gradient moment with local accumulation ( Adadelta/RMSProp)

N ( t ) = β 1 × N ( t −1) + (1 − β 1 ) × ( E ( t ) )2 , (30)
• Compute the first gradient moment

M ( t ) = β 2 × M ( t −1) + (1 − β 2 ) × E ( t ) , (31)
• Compute bias-corrected first moment and second moment estimate

N(t)
N̂(t) = , (32)
1 − β1

M(t)
M̂(t) = , (33)
1 − β2
• Update parameters

M̂(t)
W(t) = p , (34)
N̂(t) + e × I
Appl. Sci. 2019, 9, 2414 10 of 21

4. Simulation Results and Discussion

In this section, we adjust by simulations different used parameters. The beneficial use of
matrix completion in order to improve the localization accuracy is justified empirically. We also
study the recovery performance, the localization accuracy and the computational complexity of each
cited algorithm.

4.1. Determining the Best Ratio between the Number of Unknown Nodes and the Number of Anchors
It is reasonable to state that the completion of a high dimensional matrix improves the localization
accuracy. This is due to the diversity of the information introduced by each sensor node. However, from
a finite dimension, the algorithm converges and the use of additional data increases the complexity
and the execution time. To ensure the best localization accuracy and decrease the execution time of the
algorithm, we set up this ratio. Let

NbUn
ϑ= , (35)
Nb An
where NbUn is the number of unknown nodes and Nb An is the number of anchors. To determine
the best ratio ϑ in terms of localization accuracy, simulations have been done. Using 35 unknown
nodes, the number of anchors varies from 5 to 20. The localization error applying Adam respect to the
number of anchors is shown in (Figure 2). When the number of anchors reaches 10, the localization
error becomes almost stable. Figure 3 illustrates the localization error in meters depending on the
number of unknown nodes using 10 anchor nodes. It can be observed that the localization error is
almost the same from 35 unknown nodes. The trade-off between accuracy and complexity leads us to
use a ratio ϑ equal to 3.5.

3.2

2.8
Localization error (m)

2.6

2.4

2.2

1.8

1.6
5 10 15 20
Number of anchors

Figure 2. The localization error using 35 unknown nodes with the varying number of anchors in
400 m2 applying Adam for sigma shadowing = 2.

We mention that simulations have been done for each optimization algorithm and for each value
of sigma shadowing. The ratio guaranteeing the best performance is always 3.5. Thus, to simplify
presentation and without loss of generality, we present only simulation results corresponding to Adam
when working with sigma shadowing equal to 2.
Appl. Sci. 2019, 9, 2414 11 of 21

2.8

2.6

2.4

Localization error (m)

2.2

1.8

1.6
10 15 20 25 30 35 40 45 50 55
Number of unknown nodes

Figure 3. The localization error using 10 anchors with the varying number of unknown nodes in
400 m2 applying Adam for sigma shadowing = 2.

4.2. Simulation Setup

We consider a wireless sensor network of 45 sensor nodes with 10 of them being anchors and 35
being unknown nodes, placed in an area of 400 m2 (i.e., 20 m × 20 m) and its architecture is illustrated
in Figure 4. The sensor nodes (anchors and unknown nodes) are randomly placed in the studied area.

20

18

Anchors
16
Unknown nodes

14

12

10
y

0
0 2 4 6 8 10 12 14 16 18 20
x

Figure 4. Configuration of the wireless sensor network.

The accuracy of the studied algorithms is investigated over many environment realizations.
In order to simplify the presentation of the paper and without loss of generality, we present results
only for one environment test, as illustrated in Figure 4. Simulation results are consistent with other
environments. 10 RSSI measurements of prij received from sensor i (i = 1, 2, 3, ..., n) are taken at each
position j ( j = 1, 2, 3, ..., n) for each sigma shadowing value. This value is calculated in dBs as:

prij = pe − plij + aσ , (36)

where pe is the transmission power, aσ is a Gaussian random variable which describes the random
shadowing effects, plij is the pathloss calculated using Equation (2). In this paper, we use parameters
related to our laboratory: $ = 3.23, pe = 20 dBm, d0 = 1 m, f = 2.4 GHz. The sigma of the
random variable a takes the values 0, 2, 5 to study the effect of its variation on the algorithm recovery
performance and localization accuracy.
Appl. Sci. 2019, 9, 2414 12 of 21

4.3. Verification of the Low Rank Property

To apply the matrix completion technique to recover the EDM matrix from observed entries,
the matrix X̂ should have low rank r to ensure that the generated distances have a strong correlation.
We can approximate unknown entries from known ones because they are dependant. To check whether
the data matrix X̂ has a good low-rank approximation, we apply the singular value decomposition [41].
An n × n matrix X̂, can be decomposed as:

X̂ = OFQ T , (37)

where O is an n × n unitary matrix, Q is a n × n unitary matrix and F is a n × n diagonal matrix.

The diagonal elements of F are the singular values of X̂, they are organized in a decreasing order (i.e.,
F = diag(γ1 , γ2 , ..., γr , 0..., 0)). r is the rank of the matrix X̂, it is equal to the number of its non zero
singular values. If X̂ is a low rank matrix, its top l singular values occupy the total or nearly the total
energy (i.e., ∑il=1 γi2 = ∑in=1 γi2 ) [42]. The metric used to verify the property of low rank, is the fraction
of the nuclear norm caught by the top l singular values. This fraction is defined as:

∑il=1 γi2 ∑ l γ2
g( l ) = = in=1 i2 , (38)
||X̂||∗ ∑ i = 1 γi
Figure 5 illustrates the fraction of the nuclear norm captured by the top l singular values. We find
that the top 3 singular values capture 90% of the nuclear norm. This indicates that the matrix X̂ has a
low rank approximation. So, we are able to apply matrix completion.

0.95

0.9
g(l)

0.85

0.8

0.75

0.7
0 5 10 15 20 25 30 35
l

Figure 5. Energy fraction captured by the top l singular values.

4.4. Recovery Performance and Localization Accuracy

For evaluating the studied solutions based on GD and its variants combined with the trilateration
process, we define the following two metrics.

• EDM reconstruction error using the mean square error

||X̂ − Xtrue ||2F

MSE = , (39)
||Xtrue ||2F
• Localization error

||Ĉ − C||2
errorloc = , (40)
n−m
Appl. Sci. 2019, 9, 2414 13 of 21

As mentioned before, Ĉ is the matrix of estimated coordinates of unknown nodes and C are their real
coordinates.
We recall that parameters α, λ, e, µ, ρ, β 1 and β 2 are adjusted by simulations, selected to ensure
the best result in terms of localization mean error on a validation dataset. To find the best set of these
parameters, an empirical process is conducted. Thus, several simulations are required to identify the
optimal value of each parameter. The parameters used in the rest of the paper are presented in Table 2.
The rest of algorithms includes GD, NAG, Adagrad, RMSProp and Adadelta.

Table 2. Values of different used parameters.

Adam The Rest of Algorithms

α 0.5 0.5
λ 0.9 0.9
Sigma shadowing = 0 e 0.01 0.01
µ − 0.9
ρ − 0.9
β1 0.001 −
β2 0.001 −
α 0.5 0.5
λ 0.9 0.9
Sigma shadowing = 2 e 0.1 0.001
µ − 0.35
ρ − 0.35
β1 0.001 −
β2 0.001 −
α 0.5 0.5
λ 0.9 0.9
Sigma shadowing = 5 e 0.1 0.001
µ − 0.35
ρ − 0.35
β1 0.001 −
β2 0.001 −

To verify that trilateration guarantees better localization accuracy when more distance information
is provided, we firstly apply the trilateration with observed distances only. We can easily notice that
it introduces the worse localization accuracy compared to tested combinations in both noisy and
noiseless environments (Figure 6b,d,f). The localization accuracy is much better when we use a
complete EDM than using only the observed distances. Moreover, to apply the trilateration process,
at least three detected anchors are needed. If this is not the case, the sensor node cannot be localized.
This problem can be solved when using a compete EDM containing all pairwise distances. Therefore,
the combination of matrix completion technique and trilateration is highly recommended.
Appl. Sci. 2019, 9, 2414 14 of 21

1
Adam
Adadelta
10 0 RMSProp 0.9
Adagrad
NAG
GD 0.8

0.7
MSE

CDF
0.6
10 -1

0.5
Adam
Adadelta
RMSProp
0.4
Adagrad
NAG
0.3 GD
without MC
-2
10
0.2
0 2 4 6 8 10 12 0 1 2 3 4 5 6 7 8 9
Error (m)
Number of iterations 10 4

(a) EDM reconstruction error with sigma shadowing = 0 (b) Localization error with sigma shadowing = 0
1

Adam 0.9
Adadelta
10 0
RMSProp
Adagrad 0.8
NAG
GD
0.7
MSE

CDF

0.6

0.5
10 -1 Adam
0.4 Adadelta
RMSProp
Adagrad
0.3 NAG
GD
sans MC
0.2

0 2 4 6 8 10 12 0 1 2 3 4 5 6 7 8
Error (m)
number of iterations 10 4

(c) EDM reconstruction error with sigma shadowing = 2 (d) Localization error with sigma shadowing = 2

Adam 0.9
Adadelta
RMSProp 0.8
Adagrad
10 0
NAG
GD 0.7

0.6
MSE

CDF

0.5

Adam
0.4 Adadelta
RMSProp
Adagrad
0.3 NAG
GD
sans MC
0.2

-1 0.1
10
0 2 4 6 8 10 12 0 1 2 3 4 5 6 7 8 9
Error (m)
Number of iterations 10 4

(e) EDM reconstruction error with sigma shadowing = 5 (f) Localization error with sigma shadowing = 5

Figure 6. Algorithms’ performances (EDM: Euclidean distance matrix).

In the first set of simulations, we investigate the performances, in terms of EDM reconstruction and
localization error, of different cited algorithms in a noiseless environment as shown in Figure 6a,b. It
shows the effectiveness of the location estimation of the proposed schemes for 10 simulations. Figure 6a
illustrates the EDM reconstruction error ’MSE’. The error estimation of pairwise known distances
is in the order of 10−30 . The reconstruction error varies between 10−3 and 10−1 for different cited
Appl. Sci. 2019, 9, 2414 15 of 21

algorithms. The best EDM reconstruction rate is obtained by Adadelta which produces a localization
mean error of 0.47 m. This result is close to those obtained by GD, NAG and Adam. The difference is
the number of iterations required to reach the convergence of the algorithm. Adam converges at 790
1
iterations which is about the 18 of required iterations by Adadela. GD converges at 5910 iterations
which is about the 0.41 of required iterations by Adadela and NAG converges at 3100 iterations.
The performances of GD and NAG in terms of localization accuracy are very close. So that, their
cumulative distribution functions (CDF) are confused for each sigma shadowing value. However, it
accelerates the convergence of GD. Instead of converging at 5910, it converges at 3100 when working
in a noiseless indoor environment.
Adagrad exhibits the worst performances in terms of EDM reconstruction error and localization
error. This is due to the fact that it accumulates the squared gradients in the denominator. So, the sum
of positive terms keeps growing and the learning rate becomes very small, thus making the algorithm
no longer able to ensure updates in order to reach a lower minimum. RMSProp, Adadelta and Adam
aim to reduce decreasing learning rate.
Instead of accumulating all past squared gradients, RMSProp and Adadelta use a window
of size (ρ) of accumulated past gradients. RMSProp improves a little bit the EDM reconstruction
error (Figure 6a) and the localization error compared to those introduced by Adagrad in a noiseless
environment. However, the result is still worse than those obtained by GD and NAG. The performances
of RMSProp and Adagrad are quite close in a noisy environment. They converge to close minimums
(Figure 6c,e) and their CDF are almost confused (Figure 6d,f).
Adadelta performs a bit better than GD and NAG for sigma shadowing = 0 and 2. However, it
requires 2.4 times the number of iterations required by GD to converge and 4.5 times the number of
iterations required by NAG to converge (in a noiseless environment) and it is more complex than GD
and NAG, which increases the execution time of the algorithm. We notice that adadelta is more affected
by noise than other algorithms. Its performances decrease more quickly than the others. In a noiseless
indoor environment, Adadelta introduces the best localization accuracy. For sigma shadowing equal
to 2, Adam performs better than it. Furthermore, for sigma shadowing equal to 5, Adam, GD and
NAG are better than Adadelta in terms of EDM reconstruction error, localization accuracy and speed
of convergence. The advantage of Adadelta is that we do not need to set a default learning rate since
α has been eliminated from the update rule. However, this can have a negative effect as we cannot
control the learning rate. To resolve this flaw, Adam is used.
Adam works well compared to other algorithms considering the compromise between the
localization accuracy and the execution time. It requires the smallest number of iterations to converge.
This is due to the fact that it uses an accelerator term by taking into account past updates. Furthermore,
it achieves to the best localization mean error in a noisy environment. For sigma shadowing equal
to 2, its localization mean error is 1.2 m and it reaches 2.6 m when sigma shadowing is equal to
5. As mentioned before, these results have been done on 10 simulations and the variance is about
0.1 m for each value of sigma shadowing. Therefore, Adam is the most adapted algorithm for indoor
localization schemes.

4.5. Complexity

4.5.1. Analytical Expressions

This section aims at approximating the theoretical complexity of the different studied algorithms.
The complexity will be assessed by counting the number of multiplications per iteration and neglecting
additions and subtractions. The square root and the power of elements are also neglected. For matrices
multiplication, the used computational formula is the classic one and not Strassen formula [43].
To calculate the square power of a matrix, the binary method is considered [44]. The complexity of the
negative square root of a matrix is obtained by calculating the inverse of the square root of this matrix.
We calculate the complexity of the inverse of a matrix based on the Gauss method [43]. The complexity
Appl. Sci. 2019, 9, 2414 16 of 21

of the square root of a matrix is determined through the Denman-Beavers algorithm [45]. We define
cGD , c N AG , c Adagrad , c RMSProp , c Adadelta and c Adam the computational complexity per iteration of GD,
NAG, Adagrad, RMSProp, Adadelta and Adam, respectively. Furthermore, as defined before, n is the
number of sensor nodes (anchors and unknown nodes).

cGD = 4n3 + 6n2 , (41)

c N AG = 8n3 + 9n2 , (42)

c Adagrad = 5n3 + 7n2 , (43)

c RMSProp = 6n3 + 9n2 , (44)

c Adadelta = 6n3 + 10n2 , (45)

c Adam = 6n3 + 9n2 , (46)

4.5.2. Analysis
According to the closed-form expressions mentioned in Section 4.5.1, it is possible to numerically
calculate the complexity of the different algorithms , since the value of n is known, as given in Table 3.
The complexity of each iteration and the complexity of the algorithm are normalized with respect to
the complexity of GD, in order to highlight the contribution of other cited algorithms compared to GD.
The localization mean error in Table 3 corresponds to the best mean error registered by each algorithm
at different values of sigma shadowing.
The best localization mean errors reached by GD and NAG are close. However, in terms of
execution time, the GD wins. In terms of number of iterations required to converge, NAG converges
fast compared to GD. However, the complexity taken by each iteration, which represents 1.9839
operations instead of 1 operation registered by GD. The worst localization mean error is introduced
by Adagrad. It is also associated with the highest computational complexity. Its required number
of operations per iteration is the best one compared to other used algorithms (except GD). However,
the number of iterations to converge is too high, which increases significantly the complexity of the
algorithm. The complexity of RMSProp to converge is around 0.4 of the one registered by Adagrad.
However, it is still high and cannot be adapted in large networks or in real time localization systems.
The required number of operations per iteration for Adadelta and Adam is around 1.5 × cGD .
However, the number of iterations required by Adadelta to converge is significantly higher than the
number needed by Adam. For this, the execution time and the computational complexity are higher for
Adadelta than Adam. Adam corresponds to the best complexity, it is 0.2 of the complexity reached by
GD. Compared to other cited algorithms, it is associated to the lowest complexity. It does not require a
high computational time, so, it can be adapted to real time localization systems. Thus, considering
the trade-off between localization accuracy and computational complexity, Adam outperforms other
optimization algorithms.
Appl. Sci. 2019, 9, 2414 17 of 21

Table 3. Algorithms’ complexity normalized with respect to ‘Gradient descent’ (GD).

Number of Operations per Iteration

Number of Iterations to Reach the Complexity of the Algorithm to Reach the Convergence
Algorithm Localization Mean Error (m) Normalized by the Number of Operations
Convergence Respect to the Complexity of GD
per Iteration for GD
Without
matrix 4.1 — — —
completion
GD 0.56 1 5910 1
Sigma shadowing = 0 NAG 0.56 1.9839 3100 1.0406
Adagrad 1.8 1.2473 120,000 25.3261
RMSProp 1.8 1.5 43,750 11.1041
Adadelta 0.47 1.5054 14,150 3.6042
Adam 0.55 1.5 790 0.2005
Without
matrix 4.7 — — —
completion
GD 1.2 1 5535 1
Sigma shadowing = 2 NAG 1.2 1.9839 3942 1.4129
Adagrad 2.2 1.2473 125,900 28.3716
RMSProp 2.1 1.5 45,640 12.3686
Adadelta 1.2 1.5054 10,530 2.8639
Adam 1.2 1.5 775 0.21
Without
matrix 4.9 — — —
completion
GD 2.6 1 4435 1
Sigma shadowing = 5 NAG 2.6 1.9839 3281 1.4677
Adagrad 3.4 1.2473 121,900 34.2835
RMSProp 3.4 1.5 43,920 14.8546
Adadelta 2.7 1.5054 9742 3.3067
Adam 2.6 1.5 751 0.2540
Appl. Sci. 2019, 9, 2414 18 of 21

5. Method Verification Using Real Measurements

For method verification, we consider a classroom of our university as an indoor area of 88 m2
illustrated in Figure 7. The used technology is LoRa which uses 868 MHz as frequency band. We placed
eight sensor nodes which can transmit and receive messages. We collect RSSI data during an afternoon.
Then, we construct the RSSI matrix, in which there are some missing values which correspond to
unknown detected sensors. This matrix is used to obtain the partially known EDM matrix. We mention
that the used technology and the number of sensor nodes explored for experimental verification
relates to the availability of such equipment in our laboratory. Thus, we do not consider the optimal
conditions determined before. The aim is to validate the effectiveness of our method.

12,5 m

S3 S2 S1

7,6 m
S5

S4

Sensor node Si
where i = 1, 2, …,8

S8 S7 S6

10,5 m

Figure 7. Used configuration for real experimentations.

When applying our localization algorithm (Adam is used to estimate the complete EDM matrix)
to localize each sensor node, we obtain a mean localization error which is 1.5 m. This error reaches
3.8 m when using the classic trilateration. In Figure 8, we present the CDF corresponding to the classic
trilateration (without MC i.e., without matrix completion) and the CDF corresponding to our algorithm
(with MC, i.e., with matrix completion).

0.9

0.8

0.7
CDF

0.6

0.5
with MC
without MC
0.4

0.3
0 1 2 3 4 5 6 7 8 9 10 11
Error (m)

Figure 8. CDFs for real measurements.

Appl. Sci. 2019, 9, 2414 19 of 21

6. Conclusions
In this paper, we aim to improve the localization accuracy of the trilateration technique based on
pairwise distances between sensor nodes. For this, we explore the problem of missing data in sensor
network localization. Technically, we formulate the information distance from RSSI measurements
using a propagation model. We obtain a squared Euclidean distance matrix with several unknown
entries. Then, based on the matrix completion process, the complete EDM is efficiently generated by
exploring only the available distances. For this, Gradient descent based advanced methods, some of
which use an adaptative learning rate, are introduced. To perform trilateration, all pairwise distances
between the node to localize and different sensor nodes are explored. To validate the merits of
the proposed framework, an extensive set of simulations was carried out in noiseless and noisy
environments using a real propagation model. Simulation results suggest that the proposed Gradient
descent variants based matrix completion reliably estimate the complete EDM exploring partial known
information. Trilateration combined with matrix completion outperforms the traditional localization
system. Our simulation results indicate that trilateration combined with Adam which is used to solve
the matrix completion problem outperforms the other combinations. This approach does not require
a high computational complexity. Real experiments have been done in order to support simulation
results and show that our approach outperforms the classic trilateration.
Since there are pairwise distances that are better estimated than others, we are interested in using
weighted distances when applying trilateration in future work. The accorded weight increases when
the distance is well estimated. New localization system improvements in terms of complexity can be
provided using machine learning methods, since these aim to shift the prediction complexity to an
offline phase.

Author Contributions: W.N. and R.Z. conceived the presented idea and developed the closed-form of each
algorithm. Simulations and results presentation were conducted by W.N. R.Z., I.A. and M.T. verified the analytical
methods and discussed the results contributed to the final manuscript. All authors provided critical feedback and
helped the research, analysis and manuscript.
Funding: This research received no external funding.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Atzori, L.; Iera, A.; Morabito, G. The internet of things: A survey. Comput. Netw. 2010, 54, 2787–2805.
[CrossRef]
2. Hennebert, C.; Denis, B.; Le Gall, F.; Copigneaux, B.; Clari, F.; Sottile, F.; Mauro, F.; Smadja, P.; Pascali, S.;
Preuveneers, D.; et al. uBiquitous, secUre inTernet-of-things with Location and contEx-awaReness. BUTLER
project D 2, 1–134. Available online: www.iot-butler.eu (accessed on 5 July 2018).
3. Dil, B.; Dulman, S.; Havinga, P. Range-based localization in mobile sensor networks. In Proceedings of the
European Workshop on Wireless Sensor Networks, Zurich, Switzerland, 13–15 February 2006; pp. 164–179.
4. Singh, S.P.; Sharma, S.C. Range free localization techniques in wireless sensor networks: A review. Procedia
Comput. Sci. 2015, 57, 7–16. [CrossRef]
5. Awad, F.; Al-Sadi, A.; Alsmady, A. Distributed and adaptive location identification system for mobile devices.
EURASIP J. Adv. Signal Process. 2018, 2018, 61. [CrossRef]
6. Kaplan, E.; Hegarty, C. Understanding GPS: Principles and Applications; Artech House: Norwood, MA,
USA, 2005.
7. Gu, Y.; Chen, M.; Ren, F.; Li, J. HED: Handling environmental dynamics in indoor WiFi fingerprint
localization. In Proceedings of the IEEE Wireless Communications and Networking Conference (WCNC),
Doha, Qatar, 3–6 April 2016; pp. 1–6.
8. Sun, L.; Zheng, Z.W.; He, T.; Li, F. Multifloor Wi-Fi localization system with floor identification. Int. J. Distrib.
Sens. Netw. 2015, 11, 131523. [CrossRef]
Appl. Sci. 2019, 9, 2414 20 of 21

9. Cheng, J.; Song, Z.; Ye, Q.; Du, H. MIL: A mobile indoor localization scheme based on matrix completion.
In Proceedings of the IEEE International Conference on Communications (ICC), Kuala Lumpur, Malaysia,
23–27 May 2016; pp. 1–55.
10. Yang, Z.; Liu, Y. Quality of trilateration: Confidence-based iterative localization. IEEE Trans. Parallel
Distrib. Syst. 2010, 21, 631–640. [CrossRef]
11. Norrdine, A. An algebraic solution to the multilateration problem. In Proceedings of the 15th International
Conference on Indoor Positioning and Indoor Navigation, Sydney, Australia, 13–15 November 2012; p. 2012.
12. Najib Yasmeen, N.A.; Daud, N.; Abd Aziz, A.; Razali, R. 3-Dimensional wireless sensor network localization:
A review. AIP Conf. Proc. 2016, 080029. [CrossRef]
13. Kumar, P.; Reddy, L.; Varma, S. Distance measurement and error estimation scheme for RSSI based
localization in Wireless Sensor Networks. In Proceedings of the 5th IEEE Conference on Wireless
Communication and Sensor Networks (WCSN), Allahabad, India, 15–19 December 2009; pp. 1–4.
14. Nguyen, T.L.; Shin, Y. Matrix completion optimization for localization in wireless sensor networks for
intelligent IoT. Sensors 2016, 16, 722. [CrossRef] [PubMed]
15. Nguyen, L.; Kim, S.; Shim, B. Localization in Internet of Things network: Matrix completion approach.
In Proceedings of the Information Theory and Applications Workshop (ITA), San Diego, CA, USA,
31 January–2 February 2016; pp. 1–4.
16. Nguyen, L.; Shim, B. Localization of Internet of Things network via Euclidean distance matrix completion.
In Proceedings of the IEEE/CIC International Conference on Communications in China (ICCC), Chengdu,
China, 27–29 July 2016; pp. 1–4.
17. Zhu, Y.; Jiang, A.; Kwan, H.K. ADMM-Based Sensor Network Localization Using Low-Rank Approximation.
IEEE Sens. J. 2018, 18, 8463–8471. [CrossRef]
18. Alfakih, A.Y.; Khandani, A.; Wolkowicz, H. Solving Euclidean distance matrix completion problems via
semidefinite programming. Comput. Optim. Appl. 1991, 12, 13–30. [CrossRef]
19. Ferreira, R.; Xavier, J.; Costeira, J.P.; Barroso, V. Newton algorithms for Riemannian distance related problems
on connected locally symmetric manifolds. IEEE J. Sel. Top. Signal Process. 2013, 7, 634–645. [CrossRef]
20. Cai, J.-F.; Candès, E.J.; Shen, Z. A singular value thresholding algorithm for matrix completion. SIAM J. Optim.
2010, 20, 1956–1982. [CrossRef]
21. Chen, M.; Ganesh, A.; Lin, Z.; Ma, Y.; Wright, J.; Wu, L. Fast convex optimization algorithms for exact
recovery of a corrupted low-rank matrix. In Coordinated Science Laboratory Report no. UILU-ENG-09-2214;
Coordinated Science Laboratory, University of Illinois at Urbana-Champaign: Champaign, IL, USA, 2009.
22. Lin, Z.; Chen, M.; Ma, Y. The augmented lagrange multiplier method for exact recovery of corrupted
low-rank matrices. arXiv 2010, arXiv:1009.5055.
23. Liu, C.; Shan, H.; Wang, B. Wireless Sensor Network Localization via Matrix Completion Based on Bregman
Divergence. Sensors 2018, 18, 2974. [CrossRef] [PubMed]
24. Xiao, F.; Liu, W.; Li, Z.; Chen, L.; Wang, R. Noise-tolerant wireless sensor networks localization via
multinorms regularized matrix completion. IEEE Trans. Veh. Technol. 2018, 67, 2409–2419. [CrossRef]
25. Grant, M.; Boyd, S.; Ye, Y. CVX: Matlab Software for Disciplined Convex Programming; CVX Research, Inc.:
Austin, TX, USA, 2008.
26. Guo, X.; Chu, L.; Sun, X. Accurate localization of multiple sources using semidefinite programming based
on incomplete range matrix. IEEE Sens. J. 2016, 16, 5319–5324. [CrossRef]
27. Guo, X.; Chu, L.; Ansari, N. Joint localization of multiple sources from incomplete noisy Euclidean distance
matrix in wireless networks. Comput. Commun. 2018, 122, 20–29. [CrossRef]
28. Chen, J.; Mitra, U. Unimodality-Constrained Matrix Factorization for Non-Parametric Source Localization.
IEEE Trans. Signal Process. 2019, 67. [CrossRef]
29. Nguyen, L.T.; Kim, J.; Kim, S.; Shim, B. Localization of IoT Networks Via Low-Rank Matrix Completion.
IEEE Trans. Commun. 2019. [CrossRef]
30. Zhang, H.; Liu, Y.; Lei, H. Localization from Incomplete Euclidean Distance Matrix: Performance Analysis
for the SVD-MDS Approach. IEEE Trans. Signal Proces. 2019, 67. [CrossRef]
31. Torgerson, W.S. Multidimensional scaling of similarity. Psychometrika 1965, 30, 379–393. [CrossRef]
32. Tsai, M. Path-loss and shadowing (large-scale fading). Nat. Taiwan Univ. Oct. 2011, 20, 2011.
33. Recht, B.; Fazel, M.; Parrilo, P.A. Guaranteed minimum-rank solutions of linear matrix equations via nuclear
norm minimization. SIAM Rev. 2010, 52, 471–501. [CrossRef]
Appl. Sci. 2019, 9, 2414 21 of 21

34. Candès, E.J.; Recht, B. Exact matrix completion via convex optimization. Found. Comput. Math. 2009, 9, 717.
[CrossRef]
35. Duchi, J.; Hazan, E.; Singer, Y. Adaptive subgradient methods for online learning and stochastic optimization.
J. Mach. Learn. Res. 2011, 12, 2121–2159.
36. Sebastian, R. An overview of gradient descent optimization algorithms. arXiv 2016, arXiv:1609.04747.
37. Sutskever, I.; Martens, J.; Dahl, G.; Hinton, G. On the importance of initialization and momentum in
deep learning. In Proceedings of the International Conference on Machine Learning, Atlanta, GA, USA,
16–21 June 2013; pp. 1139–1147.
38. Wang, Y.; Kim, S.-P.; Principe, J.C. Comparison of TDNN training algorithms in brain machine interfaces.
In Proceedings of the IEEE International Joint Conference on Neural Networks (IJCNN’05), Montreal, QC,
Canada, 31 July–4 August 2005; pp. 2459–2462.
39. Hinton, G.; Srivastava, N.; Swersky, K. Neural networks for machine learning lecture 6a overview of
mini-batch gradient descent. Cited On 2012, 14, 31.
40. Kingma, D.P.; Ba, J. Adam: A method for stochastic optimization. arXiv 2014, arXiv:1412.6980.
41. Cheng, J.; Ye, Q.; Jiang, H.; Wang, D.; Wang, C. STCDG: An efficient data gathering algorithm based on
matrix completion for wireless sensor networks. IEEE Trans. Wirel. Commun. 2013, 12, 850–861. [CrossRef]
42. Xie, K.; Ning, X.; Wang, X.; Xie, D.; Cao, J.; Xie, G.; Wen, J. Recover corrupted data in sensor networks:
A matrix completion solution. IEEE Trans. Mob. Comput. 2017, 16, 1434–1448. [CrossRef]
43. Strassen, V. Gaussian elimination is not optimal. Numer. Math. 1969, 13, 354–356. [CrossRef]
44. Benoit, A. Algorithmique I-Cours et Travaux Dirigés L3; Ecole Normale Supérieure de Lyon: Lyon, France, 2009.
45. Iannazzo, B. On the Newton method for the matrix p th root. SIAM J. Matrix Anal. Appl. 2006, 28, 503–523.
[CrossRef]

c 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).