Adisorn Owatsiriwong
To fully understand the relationship between the neural network or deep learning and optimization problems, we can make an analogy of several technical jargon between them. Those can be summarized as follows:
Neural Network | Optimization Problems |
Epoch (one loop sweeping for all data) | Iteration |
Cross Entropy, Error function | Objective function |
Hyperparameters | Design variables, optimized variables |
Feed Forward | Compute objective value |
Back Propagation, Parameter tuning | Metaheuristics, Optimization |
Batches | Sub-iteration |
Performance curve | Convergence curve |
NN is a matrix-operated model normally used in classification, clustering, or in regression | Optimization employs the algorithms to adjust the parameters or tune the parameters in NN or DL model |
Understanding this analogy will help us grasp how both AI methods are related and can be useful in solving complex problems.
コメント