Skip to main content

Table 2 The hyperparameters of the MRL framework and their respective values selected for the current study

From: Learning regularized representations of categorically labelled surface EMG enables simultaneous and proportional myoelectric control

Hyperparameter

Symbol

Value

Floating point precision

\(b\)

\(32\) bits

Number of sEMG channels

\(I\)

\(8\)

Number of decodable DoFs

\(J\)

\(2\)

Envelope extraction filter length

\(W\)

\(0.5\) s

Size of first encoder layer

\({2}^{K}\)

\(128\)

Encoder network depth

\(N\)

\(5\)

Code size

\({2}^{K-N+1}\)

\(8\)

Decoders hidden layer size

\({2}^{S}\)

\(32\)

Contractive loss weigh

\({\alpha }_{c}\)

\({10}^{-2}\)

Adam hyperparameters

\(\upeta\)

\({10}^{-4}\)

\({\beta }_{1}\)

\(0.9\)

\({\beta }_{2}\)

\(0.999\)

Corruptive noise variance

\({\sigma }^{2}\)

\({10}^{-1}\)

Weight decay

\(\lambda\)

\({10}^{-6}\)

Minibatch size

\(B\)

\({2}^{12}\)

Validation set percentage

\(P\)

\(10\%\)

Validation lookback

\(V\)

\(300\)

Maximum number of iterations

\(M\)

\(5000\)