Fishes Go MOO: Neural Network Data Prediction

Neural Network Prediction:
NN scripts and data by N.A. Battista

Cite:
N.A. Battista, Fishes Go MOO: Pareto analysis for speed and cost of transport across a
6-dimensional design space. ______ (2025)
DOI

-----------------------------------------------------------------------------------------

MATLAB SCRIPTS:
|
|--> Prediction_LinePlots.m:
Uses the Neural Network to predict speeds across
the(f,Tamp)-subspace, ie, predicts speeds across
a 2-D slice out of the overall 6-D parameter space
|
|--> Provides plots of the speeds across particular slices.
|--> As frequency and tail beat amplitude vary, the other
4 input parameters are held constant.

|
|--> Prediction_Errors_Speed.m
Calculates the relative errors btwn the simulated speed
values and those predicted via the Neural Network
across both the training and test datasets
|
|--> Prints error statistics to the command window
|--> Provides a qualitative comparison plot
|--> Provides histograms of the PDF and CDF for the
relative errors
|
|--> Each script is self-contained. That is they contain all the
necessary supporting functions in order to run, e.g.,
forward propagation, activation function, scaling functions, etc

-----------------------------------------------------------------------------------------

DATA PROVIDED
|
|--> Trained_Neural_Network.mat
|--> Contains the trained weight matrices
and bias vectors for the NN
|
|--> TRAINING_and_TEST_data.mat
|--> Contains the input parameters and speed
values for both the training and test
datasets
|--> speed data is given in both the simulation
values as well as the transformed values
for the NN (ie, those after the Box-Cox
transforms and standardization)
|
|--> SCALING_INFO.mat
|--> Provides the data transformation parameters
for transforming the data into and out of
the NN's worldview (ie, the Box-Cox
transform params, standardization params, etc)