Table 2 DFNN model performance in Q1 with K-fold.

From: Path-based evaluation of deep learning models for solving inverse kinematics in a revolute-prismatic robot

Configuration

Architecture (hidden layers × units)

Batch size

L2 Regularization

Epoch

Cross validation loss

Single-output (θ)

4 × 128

32

0.0005

150

0.0150

Single-output (d)

0.0722

Dual-output (θ, d)

3 × 128

0.1446

Back to article page