4. Algorithm-level approach

## 4.1 Background on convolutional neural networks

## 4.1.1 Definition

F1-scores SMM recognition

> Study 1

> > S1

Time-domain

64

 CNN

Frequency-domain

Table 1. Performance

are in bold.

 rates of time-, and

frequency-domain

 CNNs for the SMM recognition

 (in terms of F1-score) and Human Activity Recognition

 CNN

96.54

 78.41

 93.62

 96.46

 95.74

 98.58

 96.07

 95.27

 85.03

 98.03

 referred to as HAR (in terms of accuracy). Highest rates

Time Series Analysis - Data, Methods, and Applications

 93.88

 93.42

 91.23

 76.76

 84.95

 93.38

 86.41

 95.11

 95.97

 75.67

 60.17

 91.68

 82.55

 84.90

99.90

95.98

 S2

 S3

 S4

 S5

 S6

 S1

 S2

 S3

 S4

 S5

 Mean

Study 2

Accuracies

 HAR

> CNNs were developed with the idea of local connectivity. Each node is connected only to a local region in the input. The local connectivity is achieved by replacing the weighted sums from the neural network with convolutions. In each layer of the CNN, the input is convolved with the weight matrix (e.g., the filter) to create a feature map. As opposed to regular neural networks, all the values in the output feature map share the same weights so that all nodes in the output detect exactly the same pattern. The local connectivity and shared weights aspect of CNNs reduce the total number of learnable parameters, resulting in more efficient training and learning in each layer a weight matrix which is capable of capturing the necessary, translation-invariant features from the input.
