This is a table of type trigram and their frequencies. Use it to search & browse the list to learn more about your study carrel.
trigram | frequency |
---|---|
convolutional neural networks | 157 |
the number of | 144 |
the performance of | 119 |
convolutional neural network | 108 |
in order to | 95 |
as well as | 92 |
based on the | 87 |
in terms of | 79 |
in this study | 76 |
of the proposed | 75 |
one of the | 75 |
performance of the | 69 |
due to the | 66 |
deep convolutional neural | 66 |
the copyright holder | 64 |
preprint in perpetuity | 63 |
is the author | 63 |
has granted medrxiv | 63 |
the preprint in | 63 |
a license to | 63 |
granted medrxiv a | 63 |
who has granted | 63 |
display the preprint | 63 |
medrxiv a license | 63 |
to display the | 63 |
the author funder | 63 |
license to display | 63 |
holder for this | 62 |
this version posted | 62 |
this preprint this | 62 |
copyright holder for | 62 |
for this preprint | 62 |
preprint this version | 62 |
of the study | 59 |
in this paper | 58 |
the use of | 58 |
detection of covid | 54 |
can be used | 52 |
of the cnn | 50 |
the proposed model | 50 |
with respect to | 49 |
used in the | 48 |
in this work | 48 |
the detection of | 48 |
deep neural networks | 47 |
the proposed cnn | 47 |
a is the | 46 |
size of the | 45 |
the proposed method | 45 |
as shown in | 45 |
it is made | 44 |
is made available | 44 |
results of the | 44 |
immune repertoire classification | 44 |
made available under | 43 |
international license it | 43 |
license it is | 43 |
a deep learning | 43 |
training and testing | 43 |
available under a | 43 |
of the model | 43 |
under a is | 43 |
modern hopfield networks | 43 |
is used to | 43 |
it can be | 42 |
the scope of | 41 |
the size of | 40 |
is shown in | 40 |
each of the | 40 |
version posted november | 39 |
within the scope | 39 |
of deep learning | 38 |
part of the | 38 |
scope of the | 38 |
according to the | 37 |
well as the | 37 |
frequency domain images | 36 |
the results of | 35 |
can be seen | 35 |
shown in figure | 35 |
the training and | 35 |
that can be | 35 |
diagnosis of covid | 35 |
used for the | 34 |
in the study | 34 |
such as the | 34 |
in this section | 33 |
the other hand | 33 |
large number of | 33 |
to reduce the | 33 |
on the other | 33 |
data with implanted | 32 |
deep learning models | 32 |
the case of | 32 |
machine learning algorithms | 32 |
to improve the | 31 |
given in table | 31 |
are given in | 31 |
is used for | 31 |
compared to the | 30 |
shown in table | 30 |
there is a | 30 |
have to be | 30 |
with implanted signals | 30 |
in the case | 29 |
patients with covid | 29 |
be used to | 28 |
as a result | 28 |
the attention mechanism | 28 |
to detect covid | 28 |
deep learning model | 28 |
test data set | 27 |
the output of | 27 |
using deep learning | 27 |
deep learning for | 27 |
neural networks for | 27 |
shown in fig | 27 |
the problem of | 27 |
we propose a | 26 |
emerson et al | 26 |
in the training | 26 |
number of instances | 26 |
the training dataset | 26 |
for the detection | 26 |
is based on | 26 |
the development of | 26 |
from chest x | 26 |
the machine learning | 26 |
be used for | 26 |
in the following | 25 |
qrelu and m | 25 |
a set of | 25 |
is one of | 25 |
machine learning methods | 25 |
in this context | 25 |
a total of | 25 |
of coronavirus disease | 25 |
for the training | 25 |
in the first | 25 |
the application of | 25 |
by peer review | 24 |
was not certified | 24 |
which was not | 24 |
a large number | 24 |
used as a | 24 |
certified by peer | 24 |
the effectiveness of | 24 |
most of the | 24 |
has to be | 24 |
in section a | 24 |
the impact of | 24 |
not certified by | 24 |
total number of | 23 |
obtained by applying | 23 |
it is possible | 23 |
the latent space | 23 |
refers to the | 23 |
of the training | 23 |
that the proposed | 23 |
of the covid | 23 |
of the images | 23 |
the diagnosis of | 23 |
of the input | 23 |
with deep convolutional | 23 |
an accuracy of | 22 |
of the most | 22 |
was used to | 22 |
the cmv dataset | 22 |
based sequence embedding | 22 |
the confusion matrix | 22 |
the proposed approach | 22 |
are shown in | 22 |
the cnn model | 22 |
to this end | 22 |
simulated immunosequencing data | 22 |
the proposed system | 22 |
a combination of | 22 |
in the second | 22 |
the learning rate | 22 |
the quality of | 22 |
of the two | 21 |
a machine learning | 21 |
the results obtained | 21 |
output of the | 21 |
the training set | 21 |
of the network | 21 |
different types of | 21 |
in the proposed | 20 |
the feature maps | 20 |
during a disaster | 20 |
is possible to | 20 |
of machine learning | 20 |
of the data | 20 |
number of parameters | 20 |
the presence of | 20 |
on social media | 20 |
some of the | 20 |
ray images using | 20 |
the same time | 20 |
the spread of | 20 |
deep learning methods | 20 |
the total number | 20 |
the purpose of | 20 |
in the future | 20 |
is given in | 20 |
p r e | 19 |
at the same | 19 |
are able to | 19 |
u r n | 19 |
as described in | 19 |
applied to the | 19 |
p r o | 19 |
evaluation of the | 19 |
multiple instance learning | 19 |
o o f | 19 |
performance of our | 19 |
n a l | 19 |
l p r | 19 |
r n a | 19 |
deep neural network | 19 |
for each of | 19 |
the experimental results | 19 |
results were obtained | 19 |
a l p | 19 |
j o u | 19 |
o u r | 19 |
in the literature | 19 |
in deep learning | 19 |
used in this | 19 |
r o o | 19 |
in the same | 19 |
classification of covid | 19 |
d cnn lstm | 18 |
the results were | 18 |
of the same | 18 |
in which the | 18 |
the training phase | 18 |
in table a | 18 |
grid search over | 18 |
the accuracy of | 18 |
the immune status | 18 |
is used as | 18 |
from patients with | 18 |
the positive class | 18 |
needs to be | 18 |
the value of | 18 |
the rest of | 18 |
the classification of | 18 |
the training data | 18 |
in the image | 18 |
the input data | 18 |
of novel coronavirus | 18 |
instances per bag | 18 |
the deep learning | 18 |
search over hyperparameters | 18 |
number of images | 18 |
which is the | 18 |
the model is | 18 |
figure shows the | 18 |
the results are | 18 |
world data with | 18 |
to evaluate the | 18 |
ray and ct | 18 |
depending on the | 17 |
there is no | 17 |
obtained from the | 17 |
support vector machines | 17 |
quality of the | 17 |
proposed cnn model | 17 |
for detection of | 17 |
of the disease | 17 |
the cnn and | 17 |
testing procedures were | 17 |
convolutional networks for | 17 |
to increase the | 17 |
extracted from the | 17 |
is able to | 17 |
neural network for | 17 |
tweets during a | 17 |
of the first | 17 |
a neural network | 17 |
ray images and | 17 |
of modern hopfield | 17 |
the importance of | 17 |
and the results | 17 |
spread of the | 17 |
respect to the | 17 |
as illustrated in | 17 |
performed using the | 17 |
and testing procedures | 17 |
due to its | 17 |
were used for | 17 |
as seen in | 17 |
neural networks and | 17 |
number of samples | 16 |
of the three | 16 |
are used for | 16 |
immunosequencing data with | 16 |
the need for | 16 |
it does not | 16 |
the end of | 16 |
of instances per | 16 |
artificial neural networks | 16 |
precision and recall | 16 |
the best results | 16 |
which is a | 16 |
consists of a | 16 |
ramsauer et al | 16 |
the qrelu and | 16 |
based on a | 16 |
values of the | 16 |
using convolutional neural | 16 |
compared to other | 16 |
recall and f | 16 |
section a we | 16 |
this work we | 16 |
support vector machine | 16 |
the effect of | 16 |
the majority of | 16 |
recurrent neural networks | 16 |
the context of | 16 |
a convolutional neural | 16 |
use of the | 16 |
be able to | 16 |
the prediction of | 16 |
neural networks with | 16 |
the lack of | 16 |
deep transfer learning | 15 |
to predict the | 15 |
with convolutional neural | 15 |
chest ct scan | 15 |
in the context | 15 |
of breast cancer | 15 |
the authors in | 15 |
be seen in | 15 |
classification accuracy of | 15 |
it is also | 15 |
of the device | 15 |
for this reason | 15 |
to fight covid | 15 |
deep learning approach | 15 |
version posted september | 15 |
be seen that | 15 |
this study is | 15 |
and machine learning | 15 |
depicted in fig | 15 |
compared with the | 15 |
the fact that | 15 |
for the purpose | 15 |
the training of | 15 |
and frequency domain | 15 |
in the next | 15 |
natural language processing | 15 |
in this case | 15 |
and can be | 15 |
nar tweets during | 15 |
of the ll | 15 |
for the first | 15 |
fully connected layers | 15 |
is applied to | 15 |
overview of the | 15 |
of our model | 15 |
be used as | 15 |
show that the | 15 |
trained on the | 15 |
an overview of | 15 |
in the input | 15 |
a variety of | 15 |
deep learning framework | 15 |
that of the | 15 |
the process of | 15 |
as a feature | 15 |
related to the | 15 |
the amount of | 15 |
information about the | 15 |
fully connected layer | 15 |
to classify the | 15 |
images of the | 15 |
acute respiratory syndrome | 15 |
is defined as | 15 |
for the classification | 15 |
the input image | 15 |
of patients with | 15 |
on computer vision | 15 |
accuracy of the | 15 |
for image classification | 15 |
the usage of | 15 |
to be used | 14 |
with regard to | 14 |
need to be | 14 |
imagenet classification with | 14 |
which can be | 14 |
for immune repertoire | 14 |
parts of the | 14 |
the test set | 14 |
of neural networks | 14 |
we used the | 14 |
the field of | 14 |
images obtained by | 14 |
showed that the | 14 |
state of the | 14 |
ml based devices | 14 |
determined by the | 14 |
deep learning algorithms | 14 |
deep learning algorithm | 14 |
with reduction to | 14 |
images using deep | 14 |
the results for | 14 |
and deep learning | 14 |
the update rule | 14 |
by using the | 14 |
structure of the | 14 |
the burden test | 14 |
the best accuracy | 14 |
computer vision and | 14 |
for training and | 14 |
were carried out | 14 |
hyperparameters with reduction | 14 |
of chest ct | 14 |
is the number | 14 |
and it is | 14 |
was applied to | 14 |
the input object | 14 |
to train a | 14 |
ray images of | 14 |
immune receptor sequences | 14 |
number of sequences | 14 |
reduction to specific | 14 |
of the image | 14 |
the svm classifier | 14 |
the result of | 14 |
the neural network | 14 |
medical image analysis | 14 |
be applied to | 14 |
conference on computer | 14 |
to the best | 14 |
it has been | 14 |
learning for image | 14 |
organized as follows | 14 |
classification with deep | 14 |
using frequency domain | 14 |
can also be | 14 |
in this way | 14 |
over hyperparameters with | 14 |
features in the | 13 |
are used to | 13 |
by applying the | 13 |
outcome of the | 13 |
the role of | 13 |
to calculate the | 13 |
on chest x | 13 |
with the proposed | 13 |
with an average | 13 |
machine learning to | 13 |
the next step | 13 |
in the third | 13 |
vision and pattern | 13 |
during the training | 13 |
and availability of | 13 |
the distance between | 13 |
proceedings of the | 13 |
analysis of the | 13 |
chest ct images | 13 |
and ct images | 13 |
an attention mechanism | 13 |
as input for | 13 |
in addition to | 13 |
object detection models | 13 |
to obtain a | 13 |
the virtual robot | 13 |
the batch size | 13 |
used as the | 13 |
networks for large | 13 |
ct scan images | 13 |
for each user | 13 |
to train the | 13 |
were used to | 13 |
to determine the | 13 |
and pattern recognition | 13 |
is composed of | 13 |
human activity recognition | 13 |
the form of | 13 |
can be defined | 13 |
to enhance the | 13 |
transfer learning with | 13 |
due to their | 13 |
in the range | 13 |
included in the | 13 |
the data is | 13 |
the range of | 13 |
the results showed | 13 |
images were used | 13 |
of the algorithm | 13 |
the type of | 13 |
image classification tasks | 13 |
severe acute respiratory | 13 |
impact of the | 13 |
the authors declare | 13 |
using deep neural | 13 |
for detecting the | 13 |
world health organization | 13 |
to the cnn | 13 |
the input of | 13 |
detection of coronavirus | 13 |
early detection of | 13 |
training and validation | 13 |
machine learning algorithm | 13 |
batch size of | 13 |
with the highest | 13 |
to each other | 13 |
the potential to | 13 |
trained on a | 13 |
of the feature | 13 |
sequences with the | 13 |
rectified linear unit | 13 |
machine learning and | 13 |
is the most | 12 |
performed on the | 12 |
on a large | 12 |
were performed using | 12 |
in the test | 12 |
present in the | 12 |
the most common | 12 |
a lot of | 12 |
been used to | 12 |
of a sequence | 12 |
be defined as | 12 |
see table a | 12 |
need and availability | 12 |
in the form | 12 |
propose a novel | 12 |
we used a | 12 |
the probability of | 12 |
terms of the | 12 |
the proportion of | 12 |
can be applied | 12 |
and analysis of | 12 |
do not have | 12 |
the hyperparameter search | 12 |
the input sequences | 12 |
full hyperparameter search | 12 |
a survey on | 12 |
of the sequences | 12 |
deep residual learning | 12 |
is that the | 12 |
to identify the | 12 |
from chest ct | 12 |
as depicted in | 12 |
procedures were performed | 12 |
is not a | 12 |
development of a | 12 |
the classification accuracy | 12 |
for machine learning | 12 |
it should be | 12 |
for image recognition | 12 |
comparison of the | 12 |
in the data | 12 |
ct images of | 12 |
the design of | 12 |
ssd mobilenet v | 12 |
rest of the | 12 |
into account the | 12 |
to extract features | 12 |
for this purpose | 12 |
marine predators algorithm | 12 |
segmentation and classification | 12 |
we do not | 12 |
modern hopfield network | 12 |
model can be | 12 |
wang et al | 12 |
of the classification | 12 |
deep convolutional networks | 12 |
seen that the | 12 |
of the virus | 12 |
of the sequence | 12 |
of convolutional neural | 12 |
during the disaster | 12 |
features from the | 12 |
of the datasets | 12 |
may not be | 12 |
the nar tweets | 12 |
taking into account | 12 |
to diagnose covid | 12 |
learning rate is | 12 |
could be used | 12 |
deep learning architectures | 12 |
in chest x | 12 |
difference between the | 12 |
as input to | 12 |
area under the | 12 |
using the pipeline | 12 |
this is a | 12 |
of this study | 12 |
dimension of the | 12 |
machine learning models | 12 |
vgg and vgg | 12 |
automated detection of | 12 |
is organized as | 12 |
in the dataset | 11 |
we present a | 11 |
social distancing monitoring | 11 |
a cnn model | 11 |
and convolutional neural | 11 |
international conference on | 11 |
detection and classification | 11 |
results for the | 11 |
cnn model was | 11 |
of which were | 11 |
to the covid | 11 |
jaccard similarity coefficient | 11 |
any of the | 11 |
in comparison to | 11 |
ml based techniques | 11 |
the structure of | 11 |
for the diagnosis | 11 |
with transfer learning | 11 |
complex wavelet transform | 11 |
to avoid the | 11 |
in the present | 11 |
to overcome the | 11 |
sequences in the | 11 |
the outcome of | 11 |
no reuse allowed | 11 |
the architecture of | 11 |
this paper is | 11 |
pipeline classification algorithms | 11 |
a transfer learning | 11 |
this is not | 11 |
learning convolutional neural | 11 |
gesture recognition system | 11 |
shown in the | 11 |
input of the | 11 |
scale image recognition | 11 |
can lead to | 11 |
we present the | 11 |
depends on the | 11 |
of ml based | 11 |
a comparison of | 11 |
is depicted in | 11 |
of the pre | 11 |
training and test | 11 |
and the m | 11 |
neural network architectures | 11 |
with rician noise | 11 |
we use the | 11 |
the choice of | 11 |
the custom model | 11 |
the pipeline classification | 11 |
for homology modelling | 11 |
than the other | 11 |
ml based systems | 11 |
automatic detection of | 11 |
the n hypothesis | 11 |
the proposed algorithm | 11 |
of a gesture | 11 |
hand gesture recognition | 11 |
the feature map | 11 |
table shows the | 11 |
the real robot | 11 |
images in the | 11 |
heartbeat segmentation and | 11 |
model is trained | 11 |
detection from x | 11 |
have been proposed | 11 |
the set of | 11 |
on machine learning | 11 |
qrelu and the | 11 |
very deep convolutional | 11 |
to the original | 11 |
residual learning for | 11 |
a feature extractor | 11 |
the details of | 11 |
allowed without permission | 11 |
has the potential | 11 |
to compute the | 11 |
reuse allowed without | 11 |
to make the | 11 |
the chest x | 11 |
better than the | 11 |
model based on | 11 |
a sequence of | 11 |
are reported in | 11 |
all rights reserved | 11 |
on the covid | 10 |
the novel coronavirus | 10 |
have shown that | 10 |
representation of the | 10 |
to that of | 10 |
fire module and | 10 |
are used as | 10 |
except for the | 10 |
can be found | 10 |
we use a | 10 |
that there is | 10 |
we show that | 10 |
of the individual | 10 |
initial learning rate | 10 |
with the following | 10 |
of the full | 10 |
to achieve a | 10 |
using ct images | 10 |
a bibliometric analysis | 10 |
are based on | 10 |
average auc of | 10 |
the treatment of | 10 |
results show that | 10 |
is given by | 10 |
in social media | 10 |
results showed that | 10 |
update rule of | 10 |
achieved the best | 10 |
the one hand | 10 |
in the real | 10 |
a deep convolutional | 10 |
the italian dataset | 10 |
and the second | 10 |
was used for | 10 |
based deep learning | 10 |
added to the | 10 |
hopfield networks and | 10 |
standard deviations across | 10 |
based on deep | 10 |
and the number | 10 |
is necessary to | 10 |
considered as a | 10 |
is statistically significant | 10 |
a result of | 10 |
used to compute | 10 |
images from the | 10 |
are presented in | 10 |
used to train | 10 |
the transfer learning | 10 |
the evaluation of | 10 |
tested on the | 10 |
given in section | 10 |
to measure the | 10 |
number of tweets | 10 |
feature vector of | 10 |
model of the | 10 |
for each genus | 10 |
paper is organized | 10 |
a validation set | 10 |
sequences per repertoire | 10 |
the model for | 10 |
have been used | 10 |
using deep convolutional | 10 |
has been used | 10 |
in table and | 10 |
features of the | 10 |
out of the | 10 |
of the ieee | 10 |
ct scans of | 10 |
the frequency domain | 10 |
the implanted motif | 10 |
feature maps from | 10 |
t cell receptor | 10 |
to be a | 10 |
feature cnns with | 10 |
that the cnn | 10 |
the full hyperparameter | 10 |
an average auc | 10 |
with the help | 10 |
the help of | 10 |
the proposed models | 10 |
it is necessary | 10 |
section presents the | 10 |
to extract the | 10 |
the d cnn | 10 |
cells and viruses | 10 |
of the implanted | 10 |
in which we | 10 |
the identification of | 10 |
input to the | 10 |
on deep learning | 10 |
early diagnosis of | 10 |
nih chest x | 10 |
cnn model is | 10 |
of the chest | 10 |
that our model | 10 |
position of the | 10 |
the dimension of | 10 |
for identifying the | 10 |
based gesture recognition | 10 |
learning framework for | 10 |
the cnn architectures | 10 |
that the attention | 10 |
and transfer learning | 10 |
in tables and | 10 |
neural networks deep | 10 |
from ct images | 10 |
of the robot | 10 |
is represented by | 10 |
the implanted signal | 10 |
positive or negative | 10 |
the production of | 10 |
a collection of | 10 |
fold cross validation | 10 |
and test set | 10 |
of ct images | 10 |
small number of | 10 |
neural network with | 10 |
availability of resource | 10 |
on the one | 10 |
an auc of | 10 |
because of the | 10 |
in the field | 10 |
sensitivity and specificity | 10 |
features such as | 10 |
according to their | 10 |
on chest ct | 10 |
to find the | 10 |
deep learning and | 10 |
the manifestos project | 10 |
seems to be | 10 |
the authors used | 10 |
focused on the | 10 |
to learn the | 10 |
models were trained | 10 |
by emerson et | 10 |
hyperparameter search of | 10 |
are indicated by | 10 |
around the world | 10 |
tensorflow object detection | 9 |
the test dataset | 9 |
the average classification | 9 |
a subset of | 9 |
the input images | 9 |
which leads to | 9 |
terms of accuracy | 9 |
spatial domain images | 9 |
based on cnn | 9 |
the risk of | 9 |
of the original | 9 |
a review of | 9 |
with resnet backbone | 9 |
the most important | 9 |
and ct scan | 9 |
automatic detection from | 9 |
of the lungs | 9 |
we want to | 9 |
the ability to | 9 |
dataset and dataset | 9 |
our model is | 9 |
in machine learning | 9 |
filters of size | 9 |
along with the | 9 |
on the mnist | 9 |
is achieved by | 9 |
was able to | 9 |
neural network and | 9 |
from spiral drawings | 9 |
used for training | 9 |
from a different | 9 |
detection of the | 9 |
ray images covid | 9 |
a series of | 9 |
the image sizes | 9 |
the early detection | 9 |
limited number of | 9 |
this type of | 9 |
ray images utilizing | 9 |
for the evaluation | 9 |
at the end | 9 |
of transfer learning | 9 |
in medical image | 9 |
ct images to | 9 |
that has been | 9 |
learning to fight | 9 |
with an accuracy | 9 |
to create a | 9 |
in combination with | 9 |
learning model for | 9 |
it is not | 9 |
the performances of | 9 |
hu et al | 9 |
learning with convolutional | 9 |
for object detection | 9 |
stochastic gradient descent | 9 |
the models are | 9 |
for the proposed | 9 |
rays and ct | 9 |
machine learning approaches | 9 |
in the diagnosis | 9 |
the capacity of | 9 |
can be detected | 9 |
utilizing transfer learning | 9 |
the area under | 9 |
considered in this | 9 |
deep learning in | 9 |
were used as | 9 |
and classification of | 9 |
majority of the | 9 |
that consists of | 9 |
the network to | 9 |
we have used | 9 |
the best performance | 9 |
a novel deep | 9 |
the network is | 9 |
spatial and frequency | 9 |
the world health | 9 |
the cnn models | 9 |
the user interface | 9 |
the raw data | 9 |
the large number | 9 |
a range of | 9 |
deep learning with | 9 |
to address the | 9 |
the distribution of | 9 |
and semantic segmentation | 9 |
of the patient | 9 |
presented in this | 9 |
followed by the | 9 |
this can be | 9 |
to get the | 9 |
to the input | 9 |
of the total | 9 |
to demonstrate the | 9 |
is due to | 9 |
ct images using | 9 |
polymerase chain reaction | 9 |
of the experiment | 9 |
object detection api | 9 |
images utilizing transfer | 9 |
the combination of | 9 |
the position of | 9 |
ieee conference on | 9 |
a method for | 9 |
is presented in | 9 |
a for the | 9 |
to the use | 9 |
to the x | 9 |
to be able | 9 |
of this paper | 9 |
for the covid | 9 |
the paper is | 9 |
to obtain the | 9 |
machine learning techniques | 9 |
detection and diagnosis | 9 |
the original image | 9 |
of the number | 9 |
seen in figure | 9 |
of the models | 9 |
the speed of | 9 |
of classification accuracy | 9 |
vaswani et al | 9 |
the pipeline algorithms | 9 |
resource tweets during | 9 |
the capability of | 9 |
case of the | 9 |
of all the | 9 |
given in fig | 9 |
by applying dt | 9 |
images used in | 9 |
description of the | 9 |
of chest x | 9 |
given to the | 9 |
artificial neural network | 9 |
were used in | 9 |
the cnn is | 9 |
conflict of interest | 9 |
properties of the | 9 |
for computer vision | 9 |
described in section | 9 |
integrated into the | 9 |
prediction of the | 9 |
pneumonia and healthy | 9 |
novel coronavirus in | 9 |
a list of | 9 |
does not have | 9 |
with deep learning | 9 |
for the early | 9 |
improve the performance | 9 |
be used in | 9 |
using the real | 9 |
defined by the | 9 |
we suggest that | 9 |
the ability of | 9 |
for the cnn | 9 |
settings of the | 8 |
tree complex wavelet | 8 |
of resource tweets | 8 |
ostmeyer et al | 8 |
the convolutional layer | 8 |
to provide a | 8 |
the area of | 8 |
with the same | 8 |
in breast cancer | 8 |
overall accuracy of | 8 |
in the two | 8 |
from cxr images | 8 |
that the model | 8 |
stressed or non | 8 |
settings used in | 8 |
note that this | 8 |
complexity of the | 8 |
of the previous | 8 |
length of the | 8 |
the vanishing gradient | 8 |
the respective value | 8 |
applying the lbp | 8 |
and for the | 8 |
the cnn architecture | 8 |
ranges are given | 8 |
the known motif | 8 |
to solve the | 8 |
italy earthquake dataset | 8 |
used in our | 8 |
tweets related to | 8 |
respective value ranges | 8 |
the range and | 8 |
for the test | 8 |
and f score | 8 |
immune status of | 8 |
details of which | 8 |
subset of the | 8 |
end of the | 8 |
with minmax kernel | 8 |
transfer learning and | 8 |
the high quality | 8 |
authors declare that | 8 |
achieved by the | 8 |
of the methods | 8 |
as part of | 8 |
a mil problem | 8 |
models can be | 8 |
our proposed multi | 8 |
cell receptor sequences | 8 |
machine learning approach | 8 |
none of the | 8 |
of patients infected | 8 |
characters in the | 8 |
to achieve this | 8 |
a tailored deep | 8 |
with each other | 8 |
in the repertoires | 8 |
lies in the | 8 |
the same accuracy | 8 |
screening of covid | 8 |
addition to the | 8 |
by the following | 8 |
in this research | 8 |
the settings of | 8 |
the proposed architecture | 8 |
in other words | 8 |
nepal and italy | 8 |
the classification performance | 8 |
a systematic review | 8 |
be considered as | 8 |
the generic models | 8 |
search as well | 8 |
in a repertoire | 8 |
the present work | 8 |
of immune repertoires | 8 |
spiral drawings benchmark | 8 |
ratio of sequences | 8 |
has also been | 8 |
in recent years | 8 |
with a small | 8 |
tail of a | 8 |
results indicated that | 8 |
for each class | 8 |
on the dataset | 8 |
set and the | 8 |
of the experiments | 8 |
automatic classification of | 8 |
distribution of the | 8 |
feature selection algorithm | 8 |
was used in | 8 |
specific number of | 8 |
machine learning in | 8 |
sequence s i | 8 |
are used in | 8 |
activation function is | 8 |
we have also | 8 |
of an image | 8 |
of images used | 8 |
as compared to | 8 |
cell receptor repertoires | 8 |
training set and | 8 |
the ll sub | 8 |
with the best | 8 |
the deeprc model | 8 |
table shows that | 8 |
of the results | 8 |
attention mechanism is | 8 |
positive class repertoires | 8 |
the nepal earthquake | 8 |
and italy earthquake | 8 |
on the imagenet | 8 |
of the four | 8 |
in the main | 8 |
to the other | 8 |
to specific number | 8 |
articulated shape model | 8 |
networks and attention | 8 |
for deep learning | 8 |
to a fixed | 8 |
object detection and | 8 |
that was used | 8 |
of this work | 8 |
the main paper | 8 |
the next section | 8 |
on top of | 8 |
is capable to | 8 |
search of the | 8 |
is performed on | 8 |
for mil problems | 8 |
the first two | 8 |
time object detection | 8 |
neural networks to | 8 |
a feature vector | 8 |
the protein sequence | 8 |
proposed model is | 8 |
cnn and lstm | 8 |
that could be | 8 |
we assume that | 8 |
found in the | 8 |
networks with x | 8 |
referred to as | 8 |
the outputs of | 8 |
and fire module | 8 |
in medical imaging | 8 |
proposed within the | 8 |
of the convolutional | 8 |
of the positive | 8 |
to select the | 8 |
the imagenet dataset | 8 |
cases using deep | 8 |
in the positive | 8 |
in the hyperparameter | 8 |
the dataset used | 8 |
the validation set | 8 |
many of the | 8 |
for cells and | 8 |
artificial intelligence and | 8 |
in the detection | 8 |
novel coronavirus disease | 8 |
neural networks are | 8 |
version posted may | 8 |
of the classifier | 8 |
to describe the | 8 |
of sequences with | 8 |
we aim to | 8 |
statistically significant difference | 8 |
were able to | 8 |
see ramsauer et | 8 |
value of the | 8 |
a uniform distribution | 8 |
in the center | 8 |
used for this | 8 |
in the sequence | 8 |
as the respective | 8 |
an fs method | 8 |
the first training | 8 |
patterns in the | 8 |
the first and | 8 |
so as to | 8 |
the ieee conference | 8 |
calculated using the | 8 |
used by cnns | 8 |
hyperparameter search as | 8 |
seen in fig | 8 |
deeprc outperforms all | 8 |
maps from the | 8 |
in the network | 8 |
the aim of | 8 |
module and fire | 8 |
transfer learning approach | 8 |
a number of | 8 |
the proposed framework | 8 |
world immunosequencing data | 8 |
found to be | 8 |
suggest that researchers | 8 |
tweets during the | 8 |
of a new | 8 |
in contrast to | 8 |
on the cmv | 8 |
the availability of | 8 |
type specific volume | 8 |
score on the | 8 |
removed from the | 8 |
table and table | 8 |
numbers of instances | 8 |
used to obtain | 8 |
rule of modern | 8 |
applied in the | 8 |
ai ml based | 8 |
to classify covid | 8 |
by machine learning | 8 |
the proposed deep | 8 |
performance of a | 8 |
the modern hopfield | 8 |
in all the | 8 |
the tail of | 8 |
the occurrence of | 8 |
method based on | 8 |
the model to | 8 |
to control the | 8 |
classification of the | 8 |
results were calculated | 8 |
the relationship between | 8 |
section describes the | 8 |
a deep neural | 8 |
the introduction of | 8 |
the witness rate | 8 |
the shallow resnet | 8 |
a for details | 8 |
value ranges are | 8 |
associated with the | 8 |
and standard deviation | 8 |
in our work | 8 |
input object x | 8 |
amount of data | 8 |
number of classes | 8 |
as training data | 8 |
to the matrix | 8 |
the second step | 8 |
cnn and knn | 8 |
depthwise separable convolutions | 8 |
reported in table | 8 |
large set of | 8 |
models using spatial | 8 |
the united states | 8 |
of the following | 8 |
so that the | 8 |
to assess the | 8 |
of mr images | 8 |
for a single | 8 |
respiratory syndrome coronavirus | 8 |
the ratio of | 8 |
proposed method is | 8 |
the provided dataset | 8 |
an input object | 8 |
acquired pneumonia and | 8 |
the attention weights | 8 |
the immune repertoire | 8 |
management of covid | 8 |
of the squeezenet | 8 |
operations to the | 7 |
proposed a novel | 7 |
leads to a | 7 |
it is important | 7 |
patients infected with | 7 |
application of deep | 7 |
also known as | 7 |
the authors have | 7 |
of the d | 7 |
the model with | 7 |
latent space is | 7 |
the possibility of | 7 |
of a deep | 7 |
feature encoder cnn | 7 |
resnet and densenet | 7 |
the neural networks | 7 |
of our proposed | 7 |
positive and negative | 7 |
the characteristics of | 7 |
to classify breast | 7 |
to detect the | 7 |
the model and | 7 |
and reduce the | 7 |
our proposed method | 7 |
learning algorithm to | 7 |
deep learning system | 7 |
drawings benchmark dataset | 7 |
spread of covid | 7 |
and the other | 7 |
impact on the | 7 |
a better understanding | 7 |
pick disease type | 7 |
data in the | 7 |
lbp and dt | 7 |
the rate of | 7 |
used to evaluate | 7 |
classification performance on | 7 |
that are correctly | 7 |
could lead to | 7 |
computer vision tasks | 7 |
of the logistic | 7 |
the frequency of | 7 |
the icu staff | 7 |
using chest x | 7 |
than that of | 7 |
validation and test | 7 |
previously described and | 7 |
how the model | 7 |
a case study | 7 |
as it is | 7 |
and nih chest | 7 |
of the method | 7 |
thanks to the | 7 |
at an early | 7 |
the concept of | 7 |
terms of classification | 7 |
the time of | 7 |
a bag of | 7 |
that they have | 7 |
in the machine | 7 |
with the other | 7 |
a pooling function | 7 |
the first step | 7 |
case of covid | 7 |
study is to | 7 |
to the training | 7 |
middle east respiratory | 7 |
was calculated for | 7 |
especially in the | 7 |
illustrated in figure | 7 |
results obtained from | 7 |
on the training | 7 |
can be stored | 7 |
tailored deep convolutional | 7 |
the convolutional layers | 7 |
assigned to the | 7 |
we did not | 7 |
which represents the | 7 |
because of its | 7 |
the second training | 7 |
and the proposed | 7 |
the extracted features | 7 |
for detecting covid | 7 |
by the total | 7 |
cnn as a | 7 |
were trained on | 7 |
resources such as | 7 |
a novel method | 7 |
the datasets in | 7 |
of data and | 7 |
for stochastic optimization | 7 |
in the experiments | 7 |
images for the | 7 |
by taking into | 7 |
in some cases | 7 |
can be calculated | 7 |
accuracy for the | 7 |
feature extraction and | 7 |
the reliability of | 7 |
proposed in the | 7 |
of a large | 7 |
the accuracy and | 7 |
shows that the | 7 |
responsible for the | 7 |
systematic review and | 7 |
the same as | 7 |
studies in which | 7 |
of the space | 7 |
for this experiment | 7 |
we can see | 7 |
network design for | 7 |
the images in | 7 |
top of the | 7 |
from the original | 7 |
of the dataset | 7 |
all of the | 7 |
to estimate the | 7 |
the training time | 7 |
was carried out | 7 |
the values of | 7 |
to produce a | 7 |
proposed quantum afs | 7 |
used as an | 7 |
corona virus disease | 7 |
and attention mechanisms | 7 |
of cnn and | 7 |
the downsampling scale | 7 |
performance on the | 7 |
with a frequency | 7 |
the learned behavior | 7 |
allows for the | 7 |
the length of | 7 |
for medical image | 7 |
similar to the | 7 |
the dnn framework | 7 |
experiments were carried | 7 |
deep learning based | 7 |
a paired t | 7 |
the sum of | 7 |
model for covid | 7 |
is given as | 7 |
declare that they | 7 |
is important to | 7 |
proposed an fs | 7 |
using a paired | 7 |
to be considered | 7 |
of the main | 7 |
they have no | 7 |
half of the | 7 |
convolution neural network | 7 |
and tested on | 7 |
disease type c | 7 |
decision support system | 7 |
carried out on | 7 |
more than one | 7 |
described and proposed | 7 |
of the different | 7 |
design for detection | 7 |
of precision and | 7 |
the effects of | 7 |
accuracy and f | 7 |
a large set | 7 |
the best result | 7 |
a kernel size | 7 |
the study results | 7 |
in latent space | 7 |
the proposed multi | 7 |
more and more | 7 |
with high accuracy | 7 |
the acquisition of | 7 |
other cnn architectures | 7 |
on the input | 7 |
a limited number | 7 |
the convolutional neural | 7 |
other machine learning | 7 |
floating point values | 7 |
image classification and | 7 |
and normal cases | 7 |
the cnn as | 7 |
generated from the | 7 |
high classification accuracy | 7 |
the single shot | 7 |
order to make | 7 |
which were previously | 7 |
even though the | 7 |
high quality range | 7 |
group within the | 7 |
within the same | 7 |
kernel size of | 7 |
sps and acp | 7 |
can be easily | 7 |
the state of | 7 |
to the total | 7 |
image obtained by | 7 |
for cnn ii | 7 |
method for stochastic | 7 |
also be used | 7 |
comparison to the | 7 |
novel deep learning | 7 |
in conjunction with | 7 |
the first cnn | 7 |
in convolutional neural | 7 |
used in many | 7 |
for the input | 7 |
were previously described | 7 |
into training and | 7 |
of cnn architectures | 7 |
and an f | 7 |
to the model | 7 |
result of the | 7 |
the weights of | 7 |
real and imaginary | 7 |
a framework for | 7 |
obtained in the | 7 |
a length of | 7 |
of a cnn | 7 |
the center of | 7 |
considered to be | 7 |
and ct scans | 7 |
at significance level | 7 |
layers of the | 7 |
chest ct findings | 7 |
classification models were | 7 |
in the last | 7 |
deep learning approaches | 7 |
are considered to | 7 |
lecun et al | 7 |
the computation of | 7 |
in real time | 7 |
test at significance | 7 |
the best of | 7 |
the issue of | 7 |
specific models using | 7 |
cnn as input | 7 |
and of the | 7 |
features extracted by | 7 |
the icu scenario | 7 |
neural network design | 7 |
large amount of | 7 |
image data collection | 7 |
the dimensions of | 7 |
the similarity between | 7 |
were obtained for | 7 |
single shot object | 7 |
the chest ct | 7 |
of the k | 7 |
this work is | 7 |
frequency of the | 7 |
number of covid | 7 |
carried out in | 7 |
the ground truth | 7 |
in the united | 7 |
the advantage of | 7 |
discussed in the | 7 |
and proposed within | 7 |
from the data | 7 |
on the performance | 7 |
chest radiography images | 7 |
of the project | 7 |
to support the | 7 |
the data set | 7 |
efficiency of the | 7 |
model was trained | 7 |
tensorflow and keras | 7 |
categories of datasets | 7 |
cnn to classify | 7 |
significant difference in | 7 |
the dataset arrangement | 7 |
spatial and temporal | 7 |
by using a | 7 |
the analysis of | 7 |
used to extract | 7 |
the images are | 7 |
a dataset of | 7 |
to the chest | 7 |
chosen to be | 7 |
transfer learning strategy | 7 |
when compared to | 7 |
to rank the | 7 |
information from the | 7 |
of an individual | 7 |
nepal earthquake dataset | 7 |
parameters of the | 7 |
experimental group within | 7 |
presented in section | 7 |
of sequence space | 7 |
the softmax function | 7 |
a psnr of | 7 |
higher than the | 7 |
are responsible for | 7 |
of the signal | 7 |
to prevent the | 7 |
in such a | 7 |
east respiratory syndrome | 7 |
were calculated using | 7 |
number of layers | 7 |
is treated as | 7 |
cases from chest | 7 |
ray and nih | 6 |
the rician noise | 6 |
the fully connected | 6 |
and red indicating | 6 |
the depth of | 6 |
for the second | 6 |
a large amount | 6 |
with blue indicating | 6 |
and diagnosis of | 6 |
the model was | 6 |
tasks such as | 6 |
of images that | 6 |
are combined to | 6 |
to have the | 6 |
from the covid | 6 |
inception architecture for | 6 |
was performed on | 6 |
nature remains neutral | 6 |
of this research | 6 |
implanted with a | 6 |
effect on the | 6 |
with bibliometric analysis | 6 |
best of our | 6 |
the same number | 6 |
maps and institutional | 6 |
from the two | 6 |
the comparison of | 6 |
resnet and the | 6 |
claims in published | 6 |
the activations are | 6 |
seen in table | 6 |
a modern hopfield | 6 |
length d v | 6 |
local binary pattern | 6 |
to complex cholesterol | 6 |
mnist benchmark dataset | 6 |
deep repertoire classification | 6 |
was used as | 6 |
weights of the | 6 |
characters with probability | 6 |
a shallow cnn | 6 |
neutral with regard | 6 |
paper is structured | 6 |
towards the prediction | 6 |
on a validation | 6 |
not able to | 6 |
data of the | 6 |
in which they | 6 |
root mean square | 6 |
for depression detection | 6 |
sundararajan et al | 6 |
d r r | 6 |
images from patients | 6 |
repertoires of the | 6 |
classification based on | 6 |
by cnn models | 6 |
a probability of | 6 |
apply ig to | 6 |
obtained for the | 6 |
was chosen to | 6 |
maps in the | 6 |
the other models | 6 |
used to calculate | 6 |
fraction of sequences | 6 |
is described in | 6 |
in the event | 6 |
of sequences in | 6 |
as an alternative | 6 |
to the corresponding | 6 |
bacterial pneumonia and | 6 |
note springer nature | 6 |
over cv folds | 6 |
taught and untaught | 6 |
each input object | 6 |
and institutional affiliations | 6 |
a survey of | 6 |
are standard deviations | 6 |
trained and tested | 6 |
and false negative | 6 |
layer is used | 6 |
indicating negative contribution | 6 |
to permutations of | 6 |
to capture the | 6 |
and compared with | 6 |
experiments are performed | 6 |
it is the | 6 |
by the world | 6 |
of feature maps | 6 |
summarized in table | 6 |
models have been | 6 |
they found that | 6 |
the aa motif | 6 |
would like to | 6 |
to screen for | 6 |
the leaky relu | 6 |
classifiers such as | 6 |
feature selection algorithms | 6 |
to the relu | 6 |
and optimized by | 6 |
sequence in the | 6 |
cwt to the | 6 |
springer nature remains | 6 |
to be the | 6 |
a grid search | 6 |
the gold standard | 6 |
volkens et al | 6 |
of medical images | 6 |
details on the | 6 |
blue indicating positive | 6 |
as an input | 6 |
invariant to permutations | 6 |
features of patients | 6 |
results in the | 6 |
to the required | 6 |
model with the | 6 |
study is the | 6 |
feature maps are | 6 |
of an input | 6 |
importance of the | 6 |
relevant for the | 6 |
drawn from the | 6 |
is likely to | 6 |
cwt operations to | 6 |
the inception architecture | 6 |
extract features from | 6 |
the work of | 6 |
and on the | 6 |
images that are | 6 |
is used in | 6 |
stacked convolutional neural | 6 |
defined as follows | 6 |
clinical features of | 6 |
survey on deep | 6 |
the visualization of | 6 |
similarity coefficient of | 6 |
be found in | 6 |
leading to the | 6 |
to the fact | 6 |
corresponds to the | 6 |
for the development | 6 |
we report the | 6 |
and accurate diagnosis | 6 |
model to classify | 6 |
combined to the | 6 |
bit floating point | 6 |
which means that | 6 |
the search space | 6 |
indicating positive contribution | 6 |
inner validation set | 6 |
propose a new | 6 |
an open source | 6 |
have the same | 6 |
errors are standard | 6 |
removed by d | 6 |
shot object detectors | 6 |
a single ct | 6 |
the mr images | 6 |
our cnn model | 6 |
been applied to | 6 |
we have shown | 6 |
the two datasets | 6 |
to build a | 6 |
defined as the | 6 |
to the meta | 6 |
motifs in the | 6 |
summarized as follows | 6 |
as a consequence | 6 |
the experiments are | 6 |
deep learning architecture | 6 |
and or healthy | 6 |
vectors are combined | 6 |
mers or sequences | 6 |
then used to | 6 |
cnn and expert | 6 |
the sequence length | 6 |
the implanted motifs | 6 |
weber et al | 6 |
the highest accuracy | 6 |
we added a | 6 |
shown to be | 6 |
training of the | 6 |
single ct image | 6 |
for classification of | 6 |
the feature cube | 6 |
to be performed | 6 |
learning system to | 6 |
models using frequency | 6 |
to provide the | 6 |
resulting in a | 6 |
d cnn kernels | 6 |
the data used | 6 |
each sequence in | 6 |
the proposed denoising | 6 |
wise relevance propagation | 6 |
is an instance | 6 |
data used in | 6 |
tool for the | 6 |
models such as | 6 |
an example of | 6 |
there are many | 6 |
not have a | 6 |
a stride of | 6 |
pick and place | 6 |
images of covid | 6 |
jurisdictional claims in | 6 |
an overall accuracy | 6 |
intel core i | 6 |
image can be | 6 |
under the roc | 6 |
and their corresponding | 6 |
proposed deep learning | 6 |
remains neutral with | 6 |
detecting the nar | 6 |
the average accuracy | 6 |
the event of | 6 |
this kind of | 6 |
the lbp and | 6 |
the infectivity of | 6 |
as feature extractors | 6 |
we show the | 6 |
mr image reconstruction | 6 |
of the learning | 6 |
from social media | 6 |
of the networks | 6 |
contribution towards the | 6 |
represents the number | 6 |
the authors also | 6 |
exponentially many patterns | 6 |
region based object | 6 |
this experiment were | 6 |
with novel coronavirus | 6 |
learning in medical | 6 |
imagenet large scale | 6 |
compared with other | 6 |
logistic regression model | 6 |
number of trials | 6 |
the above equation | 6 |
of the studies | 6 |
the most popular | 6 |
column of row | 6 |
were collected from | 6 |
is trained on | 6 |
and object detection | 6 |
the reported errors | 6 |
performance over cv | 6 |
detected by model | 6 |
as discussed in | 6 |
of the neural | 6 |
gesture recognition using | 6 |
in the latent | 6 |
no conflict of | 6 |
the difference between | 6 |
the zhao dataset | 6 |
and imaginary parts | 6 |
the diagnosis and | 6 |
visualization of the | 6 |
the success of | 6 |
machine learning model | 6 |
early detection and | 6 |
using spatial images | 6 |
in the fourth | 6 |
into consideration the | 6 |
able to detect | 6 |
data set and | 6 |
is optimized by | 6 |
at the first | 6 |
such as age | 6 |
availability of resources | 6 |
of our knowledge | 6 |
and management of | 6 |
to the negative | 6 |
the real and | 6 |
for feature extraction | 6 |
indicate higher contribution | 6 |
of attention heads | 6 |
performance evaluation of | 6 |
probability of being | 6 |
of sequences per | 6 |
low witness rate | 6 |
regard to jurisdictional | 6 |
scale visual recognition | 6 |
the results indicated | 6 |
creative commons license | 6 |
a deep cnn | 6 |
there are two | 6 |
image sizes given | 6 |
it is worth | 6 |
r d r | 6 |
the roc curve | 6 |
the jaccard kernel | 6 |
used for image | 6 |
the compared methods | 6 |
for the lstm | 6 |
the models were | 6 |
the percentage of | 6 |
using the adam | 6 |
be noted that | 6 |
the model performance | 6 |
it is a | 6 |
detection from chest | 6 |
than of the | 6 |
results are reported | 6 |
deeper with convolutions | 6 |
number of cases | 6 |
as the hyperparameter | 6 |
the following thresholds | 6 |
on the inner | 6 |
for the treatment | 6 |
are given as | 6 |
the adam optimizer | 6 |
the rectified linear | 6 |
the emergence of | 6 |
that it is | 6 |
representation of a | 6 |
greiff et al | 6 |
the computation time | 6 |
model and the | 6 |
to overcome this | 6 |
the hyperparameter and | 6 |
classification and segmentation | 6 |
the middle of | 6 |
for immune receptor | 6 |
sizes given to | 6 |
b cell receptor | 6 |
receptor sequences and | 6 |
to the number | 6 |
of the paper | 6 |
depicted in figure | 6 |
to jurisdictional claims | 6 |
cnn architecture for | 6 |
breast cancer using | 6 |
of the expert | 6 |
in the model | 6 |
larger characters in | 6 |
studies have been | 6 |
in the mr | 6 |
stressed and non | 6 |
on the number | 6 |
of using the | 6 |
the spatial domain | 6 |
imaginary parts of | 6 |
used to detect | 6 |
on convolutional neural | 6 |
found that the | 6 |
a value of | 6 |
experimental group are | 6 |
pneumonia and normal | 6 |
focus on the | 6 |
layer of the | 6 |
which in turn | 6 |
storage capacity of | 6 |
in the early | 6 |
generative adversarial networks | 6 |
the user can | 6 |
all the experiments | 6 |
the model of | 6 |
the course of | 6 |
classification accuracy and | 6 |
to investigate the | 6 |
the kernel size | 6 |
treated as the | 6 |
the results show | 6 |
we apply ig | 6 |
better understanding of | 6 |
evaluate the performance | 6 |
outperforms all other | 6 |
observed immune receptor | 6 |
incorrectly classified as | 6 |
published maps and | 6 |
the loss function | 6 |
been carried out | 6 |
have used the | 6 |
brown et al | 6 |
an acc and | 6 |
not included in | 6 |
heartbeat context information | 6 |
standard deviation of | 6 |
terms of auc | 6 |
are as follows | 6 |
average performance over | 6 |
only a few | 6 |
and clinical effects | 6 |
order to improve | 6 |
followed by a | 6 |
category simulated immunosequencing | 6 |
was set to | 6 |
we applied the | 6 |
are applied to | 6 |
the need and | 6 |
is expected to | 6 |
deep feature extraction | 6 |
particle swarm optimization | 6 |
the transformer attention | 6 |
used to generate | 6 |
vector of length | 6 |
for a specific | 6 |
we note that | 6 |
all other methods | 6 |
visual recognition challenge | 6 |
first and second | 6 |
both datasets are | 6 |
patterns from spiral | 6 |
small fraction of | 6 |
combined with the | 6 |
the lstm layer | 6 |
group are given | 6 |
to the current | 6 |
neural networks on | 6 |
the prey and | 6 |
motivated by the | 6 |
a change in | 6 |
network can be | 6 |
in the dimension | 6 |
neural network architecture | 6 |
rectified linear units | 6 |
indicated by z | 6 |
the sake of | 6 |
contribution and red | 6 |
to use the | 6 |
on the image | 6 |
going deeper with | 6 |
can readily be | 6 |
a virtual robot | 6 |
estimation of the | 6 |
the mnist benchmark | 6 |
and analyze the | 6 |
of t cell | 6 |
range of the | 6 |
extremely low witness | 6 |
patterns that can | 6 |
bigru and cnn | 6 |
be formulated as | 6 |
they do not | 6 |
the solved structure | 6 |
a light cnn | 6 |
to the user | 6 |
of cnn models | 6 |
be described as | 6 |
shows examples of | 6 |
of the diseased | 6 |
reconstruction of mr | 6 |
repertoires with unknown | 6 |
place of cnn | 6 |
attention mechanism and | 6 |
for the sake | 6 |
the results in | 6 |
in the images | 6 |
first cnn architecture | 6 |
of being removed | 6 |
based on this | 6 |
from the trained | 6 |
in our study | 6 |
in deep neural | 6 |
in patients with | 6 |
and the rest | 6 |
this applies to | 6 |
exponential storage capacity | 6 |
chest ct for | 6 |
to remove the | 6 |
from the experimental | 6 |
of this approach | 6 |
of the hand | 6 |
architecture for computer | 6 |
is trained using | 6 |
for sequence embedding | 6 |
decision tree classifier | 6 |
with depthwise separable | 6 |
on the same | 6 |
number of attention | 6 |
the training process | 6 |
between the two | 6 |
in the above | 6 |
rethinking the inception | 6 |
to train and | 6 |
fed into the | 6 |
packaging type specific | 6 |
models on the | 6 |
to test the | 6 |
of cnn kernels | 6 |
coding of the | 6 |
to simulate the | 6 |
millions of sequences | 6 |
transformer attention mechanism | 6 |
l r d | 6 |
is responsible for | 6 |
to a certain | 6 |
is in the | 6 |
axial brain image | 6 |
cnn and cnn | 6 |
like attention mechanism | 6 |
in the cnn | 6 |
montavon et al | 6 |
to the same | 6 |
the level of | 6 |
we found that | 6 |
input data size | 6 |
the applicability of | 6 |
of artificial intelligence | 6 |
r a r | 6 |
degree of substitution | 6 |
cnn model to | 6 |
used as input | 6 |
is less than | 6 |
in which this | 6 |
speed up the | 6 |
this study was | 6 |
do not use | 6 |
in the authors | 6 |
the capacity to | 6 |
and fully connected | 6 |
has been shown | 6 |
the training sets | 6 |
we propose to | 6 |
is associated with | 6 |
other methods with | 6 |
approach can be | 6 |
the space of | 6 |
hybrid deep learning | 6 |
for the identification | 6 |
be integrated into | 6 |
images and deep | 6 |
cnn is used | 6 |
range and doppler | 6 |
hopfield networks with | 6 |
with probability of | 6 |
to ensure that | 6 |
resolution of the | 6 |
for the relative | 6 |
survey with bibliometric | 6 |
for a better | 6 |
vary the number | 6 |
in clinical practice | 6 |
of our system | 6 |
the idea of | 6 |
net with resnet | 6 |
figure shows examples | 6 |
the predictive performance | 6 |
identification of covid | 6 |
in proposed a | 6 |
chest ct scans | 6 |
was trained on | 6 |
svm with minmax | 6 |
negative contribution towards | 6 |
limited availability of | 6 |
size of x | 6 |
reported errors are | 6 |
may be used | 6 |
a cnn is | 6 |
for this work | 6 |
red indicating negative | 6 |
in published maps | 6 |
calculated for the | 6 |
of deep neural | 6 |
being removed by | 6 |
the coronavirus disease | 6 |
results with the | 6 |
the extraction of | 6 |
an immune repertoire | 6 |
with simple bypass | 6 |
input for this | 6 |
for diagnosing covid | 6 |
r r r | 6 |
models for the | 6 |
patients from chest | 6 |
positive contribution and | 6 |
hyperparameter and optimized | 6 |
large scale visual | 6 |
classify breast cancers | 6 |
same number of | 6 |
cell receptor repertoire | 6 |
of the performance | 6 |
generic models using | 6 |
there are some | 6 |
l and l | 6 |
models are trained | 5 |
updates of the | 5 |
value was calculated | 5 |
because it is | 5 |
fusion operation compared | 5 |
we consider a | 5 |
proposed method with | 5 |
a sensitivity of | 5 |
is worth noting | 5 |
the early diagnosis | 5 |
in a sequence | 5 |
of our approach | 5 |
the imaginary part | 5 |
framework for the | 5 |
experiments are conducted | 5 |
the world and | 5 |
the original images | 5 |
the experimental group | 5 |
more information about | 5 |
by reducing the | 5 |
recorded data by | 5 |
the two novel | 5 |
statistically significant compared | 5 |
middle of the | 5 |
it is used | 5 |
and support vector | 5 |
to be demonstrated | 5 |
which have been | 5 |
based approach to | 5 |
and the best | 5 |
the intensive care | 5 |
two types of | 5 |
the first level | 5 |
with a large | 5 |
in figures and | 5 |
imaginary part of | 5 |
a measure of | 5 |
feature extractor and | 5 |
of feature detectors | 5 |
terms of area | 5 |
to the real | 5 |
lin et al | 5 |
the task of | 5 |
comparison to rt | 5 |
fake and spam | 5 |
the other algorithms | 5 |
and then the | 5 |
difference in the | 5 |
then applied to | 5 |
they are not | 5 |
the product name | 5 |
the response of | 5 |
structures of the | 5 |
information in a | 5 |
the authors propose | 5 |
the department of | 5 |
should not be | 5 |
the second best | 5 |
of their respective | 5 |
scale feature maps | 5 |
to be identified | 5 |
images based on | 5 |
of covid from | 5 |
model to learn | 5 |
used to classify | 5 |
klambauer et al | 5 |
method does not | 5 |
has been trained | 5 |
suggest that the | 5 |
for human activity | 5 |
for the prediction | 5 |
explicit heartbeat segmentation | 5 |
of the catheter | 5 |
rest of this | 5 |
statistical significances between | 5 |
of a novel | 5 |
of the vector | 5 |
regarded as a | 5 |
recognition based on | 5 |
as the training | 5 |
cnn and the | 5 |
networks deep learning | 5 |
they can be | 5 |
in tensorflow and | 5 |
and a sensitivity | 5 |
an input image | 5 |
is considered to | 5 |
tasks involved in | 5 |
such a way | 5 |
as an example | 5 |
two novel quantum | 5 |
a morphing behavior | 5 |
the uci spiral | 5 |
to be evaluated | 5 |
noise in the | 5 |
learning model to | 5 |
a medical device | 5 |
of the medical | 5 |
compared with mono | 5 |
performance of efficientnet | 5 |
of the time | 5 |
set of n | 5 |
indicates that the | 5 |
are extracted from | 5 |
into the cnn | 5 |
to the device | 5 |
the images were | 5 |
as a pre | 5 |
by considering the | 5 |
crucial role in | 5 |
proposed approach is | 5 |
that there are | 5 |
probability of a | 5 |
to use a | 5 |
bibliometric analysis of | 5 |
as given in | 5 |
recent advances in | 5 |
the feature vector | 5 |
is characterized by | 5 |
proposed cnn based | 5 |
based on convolutional | 5 |
neural networks by | 5 |
receiver operating characteristic | 5 |
proportion of the | 5 |
during the lifetime | 5 |
been proposed to | 5 |
with the feature | 5 |
is important for | 5 |
the decision tree | 5 |
the proposed qrelu | 5 |
in most cases | 5 |
accuracy of cnn | 5 |
can be more | 5 |
as the jaccard | 5 |
with a length | 5 |
by the user | 5 |
even with a | 5 |
for a given | 5 |
trend of covid | 5 |
of cds to | 5 |
for a more | 5 |
fed to the | 5 |
a crucial role | 5 |
from the input | 5 |
for this study | 5 |
for diagnosis of | 5 |
table presents the | 5 |
class instance segmentation | 5 |
a more general | 5 |
with a higher | 5 |
ml based algorithms | 5 |
the discriminant features | 5 |
in a more | 5 |
supervised machine learning | 5 |
national institutes of | 5 |
a batch size | 5 |
architecture for the | 5 |
the images was | 5 |
transfer learning to | 5 |
a model is | 5 |
accept the n | 5 |
on the lungs | 5 |
high undersampling rates | 5 |
m is the | 5 |
with an input | 5 |
implementation of the | 5 |
the threshold value | 5 |
of the real | 5 |
diagram of the | 5 |
an activation function | 5 |
a mixed reality | 5 |
a confusion matrix | 5 |
strong evidence that | 5 |
i represents the | 5 |
press briefings corpus | 5 |
applied machine learning | 5 |
as mentioned in | 5 |
which this value | 5 |
is obtained by | 5 |
a large dataset | 5 |
understanding of the | 5 |
can be controlled | 5 |
cnn models and | 5 |
between template and | 5 |
which have a | 5 |
in the normal | 5 |
generated by the | 5 |
and sensitivity recall | 5 |
difference in terms | 5 |
the early stopping | 5 |
such as chest | 5 |
deep learning techniques | 5 |
pneumonia in wuhan | 5 |
the original features | 5 |
the training loss | 5 |
to our knowledge | 5 |
advances in the | 5 |
trained from scratch | 5 |
data with the | 5 |
analysis of covid | 5 |
and deep transfer | 5 |
we aimed at | 5 |
to carry out | 5 |
by the model | 5 |
qrelu or m | 5 |
were removed from | 5 |
reverse transcriptase for | 5 |
and artificial intelligence | 5 |
is to learn | 5 |
of features in | 5 |
order to reduce | 5 |
there has been | 5 |
diagnoses of covid | 5 |
the third and | 5 |
cases that are | 5 |
of the pipeline | 5 |
by means of | 5 |
vanishing gradient problem | 5 |
in figure a | 5 |
be adapted to | 5 |
fischer et al | 5 |
to show the | 5 |
as a classifier | 5 |
the object recognition | 5 |
authors in proposed | 5 |
images to screen | 5 |
ray images with | 5 |
outside of the | 5 |
has been widely | 5 |
we believe that | 5 |
to monitor the | 5 |
the predictions of | 5 |
caused by a | 5 |
recurrent neural network | 5 |
real part of | 5 |
the samples dataset | 5 |
used for feature | 5 |
both disaster datasets | 5 |
larger number of | 5 |
for disease control | 5 |
to be integrated | 5 |
x fewer parameters | 5 |
infected patients using | 5 |
the tweets related | 5 |
a single image | 5 |
everingham et al | 5 |
patient with covid | 5 |
is known as | 5 |
is trained to | 5 |
significant difference between | 5 |
the bottle type | 5 |
effects of the | 5 |
medical image classification | 5 |
have proposed a | 5 |
mass video surveillance | 5 |
significant compared to | 5 |
of area and | 5 |
on the test | 5 |
ct for covid | 5 |
the current situation | 5 |
cannot be applied | 5 |
the reconstruction of | 5 |
improving neural networks | 5 |
located at the | 5 |
and future research | 5 |
test data sets | 5 |
the detection and | 5 |
for corona virus | 5 |
measures for the | 5 |
are summarized as | 5 |
as active drugs | 5 |
th and th | 5 |
the th and | 5 |
it is evident | 5 |
algorithm using ct | 5 |
does not require | 5 |
the models for | 5 |
the experiments were | 5 |
the kaggle spiral | 5 |
the generation of | 5 |
images of lungs | 5 |
batch size is | 5 |
is that it | 5 |
carried out using | 5 |
a binary classification | 5 |
stages of the | 5 |
such as those | 5 |
image classification in | 5 |
second cnn architecture | 5 |
area and morphology | 5 |
as its input | 5 |
important for the | 5 |
local phase image | 5 |
the parameters of | 5 |
the positive or | 5 |
annotated political manifestos | 5 |
institutes of health | 5 |
statistically significant than | 5 |
classification as well | 5 |
the data and | 5 |
the optimization problem | 5 |
could be achieved | 5 |
our proposed system | 5 |
from the model | 5 |
evidence that the | 5 |
the lifetime of | 5 |
is a non | 5 |
in this dataset | 5 |
for the case | 5 |
and viral pneumonia | 5 |
different numbers of | 5 |
results in a | 5 |
feature cnn architectures | 5 |
features are extracted | 5 |
classification using deep | 5 |
robot can be | 5 |
are the most | 5 |
be the case | 5 |
the adoption of | 5 |
with gb ram | 5 |
respiratory distress syndrome | 5 |
better results than | 5 |
space of the | 5 |
evaluate the results | 5 |
fs method based | 5 |
of sequences for | 5 |
improve the classification | 5 |
is calculated as | 5 |
the immune system | 5 |
clinical characteristics of | 5 |
first case of | 5 |
the disease and | 5 |
we have proposed | 5 |
tool for covid | 5 |
models used in | 5 |
cnn model for | 5 |
with more than | 5 |
the pooling layer | 5 |
o and o | 5 |
to the nearest | 5 |
learning approach for | 5 |
this dataset is | 5 |
allows us to | 5 |
results in terms | 5 |
to cope with | 5 |
algorithms based on | 5 |
to the next | 5 |
on the validation | 5 |
based convolutional neural | 5 |
the power of | 5 |
been shown to | 5 |
approach is to | 5 |
feature maps of | 5 |
to detect depression | 5 |
we used two | 5 |
neural networks a | 5 |
corrupted with rician | 5 |
models to detect | 5 |
is an important | 5 |
discriminate between covid | 5 |
means that the | 5 |
the backbone network | 5 |
the degree of | 5 |
covid from radiographs | 5 |
as a medical | 5 |
three types of | 5 |
in this regard | 5 |
in the dnn | 5 |
identifying the nar | 5 |
framework for screening | 5 |
the same way | 5 |
the most discriminating | 5 |
control the spread | 5 |
of the last | 5 |
in coronavirus disease | 5 |
up to now | 5 |
feature selection method | 5 |
as it does | 5 |
when using the | 5 |
to ensure the | 5 |
of data in | 5 |
study on the | 5 |
from a single | 5 |
we demonstrate that | 5 |
is that they | 5 |
feature maps to | 5 |
during the covid | 5 |
basic coding of | 5 |
divided into two | 5 |
according to a | 5 |