|
| Algorithms | Advantages | Disadvantages |
|
| Radiomics [274] | It could extricate and distinguish many provisions and component types. It has a minimal expense. | For respiratory movement, it obscures data. It has restricted data of remade pictures. |
| Transfer learning and fine-tuning [244] | It saves preparation time, does better execution of neural organizations, does not require a great deal of data, and can prompt higher exactness | Transfer learning has the issue of negative exchange. Fine-tuning can at some point befuddle to sort out subclasses. |
| LSTM + CNN [94] | It is appropriate to separate compelling elements and group, process, and foresee time series given delays of obscure length | It is inclined to overfitting, and it is hard to apply as it requires 4 direct layers which require a lot of memory |
| Standard deviation [275] | It gives an exact thought of how the data are appropriated. It is detached by outrageous qualities. | It tends to be affected by anomalies, is hard to ascertain or comprehend, and works out all vulnerability as error |
| Autoencoder [276] | It can be incredible for highlight extraction, conservativeness, and speed in coding utilizing backpropagation | It cannot deal with adequate preparation information, prepares some unacceptable use cases, and is excessively lossy |
| Variance [277] | It treats all deviations from the mean and assists an association with being proactive in accomplishing targets | It gives added weight to anomalies, is not effectively deciphered, and does not offer wonderful precision |
| Fourth-moment kurtosis [50] | It will be in the positive structure, and conveyance about the mean gets tighter as the mean gets bigger | The weakness is that it will not have a negative or indistinct structure |
| Wavelet [278] | It offers a synchronous restriction on schedule and recurrence space. It is quick and can isolate the fine subtleties in a sign. | It has shift affectability, its directionality is poor, and it has absence of stage data |
| Entropy [279] | It is utilized in any course of weight assurance. It is vigorous and computationally basic. | It has restricted critical thinking part and relative disparity, contingent upon the given length and biasing |
| Histogram of oriented gradients [267] | It shows invariance to photometric changes by making a dark foundation with white molecules which sharpens the articles unmistakably | The last descriptor vector develops bigger to set more effort to extricate and to prepare utilizing a given classifier |
| Third-moment skewness [50] | It is smarter to gauge the presentation of the speculation returns, transforming the data point of high skewness into slanted conveyance | It is eccentric. The ascent and defeat of a network are best instances of the skewness. |
| AlexNet, VGG16, and VGG19 [280] | AlexNet has 8 layers that exceed the yield dissimilar to other enactment capacities. VGG is an incredible structure block for learning reasons. | AlexNet battles to examine all provisions accordingly delivering helpless performing models. VGGNet is agonizing to prepare and its loads itself are very huge. |
|
