An extended abstract for the talk given on this work at the 2023 Fourth Research Workshop on the Rapid Estimation of Fish Age Using Fourier Transform Near Infrared Spectroscopy is here:

Proceedings of the Fourth Research Workshop on the Rapid Estimation of Fish Age Using Fourier Transform Near Infrared Spectroscopy. 2024, Editors: Matta, M. E. (editors); Helser, T. E. AFSC processed report; 2024-01. DOI: https://doi.org/10.25923/g30w-8268. Follow this link for the presentation slide deck.

The Groundfish Subcommittee, a part of the Scientific and Statistical Committee (SSC) of the Pacific Fishery Management Council, reveiwed these methods in early Oct 2024. The slide deck also includes a results section that was not presented due to lack of time.

Information on Pacific hake and the Rougheye/Blackspotted rockfish Complex (with additional slides by Jim Hastie) was also presented.

This slide deck is a look at the Sablefish reference scan data.

---

Neural Net Models using the keras R Package with a Custom TensorFlow Conda Environment in Windows

Neural Net Model Code in R

The Hake scripts under 'R_NN_Model_Hake_Scripts' are cleaned up. All other species to be looked at, including Sablefish, will start will this code with the species changed. A function taking a species name would be possible down the road.

Setting Up the TensorFlow Conda Environment under Windows

See the code in 'Setting_up_TensorFlow_Conda_Environment_under_Windows' to setup the custom TensorFlow Conda environment. Note that as of March 2023, the keras R package's ability to create its own TensorFlow Conda environment is currently broken due to incompatible versions of supporting software in the environment setup.

A TensorFlow Conda environment setup under Windows 10 also works under Windows 11 and can be copied and shared (zip first since there are a lot of small files). However, the Windows 10 Conda environment does not work in R on a client being served by Windows Server 2019, but I do include a partial solution in 'Install TensorFlow on Win Server 2019 Conda Env.md' for R and another that works fully under Python with GPU support. A conda Keras install would be needed inside the environmnent to run Keras under Python, see: https://www.activestate.com/resources/quick-reads/what-is-a-keras-model/

---

Notes for the readSpectraData() Function

The directory structure used by readSpectraData() is for the R working directory to be directly below the current directory:

 ../R 

with the spectra data to be beside it in:

../OPUS Spectra

and the downloads for the data to be in this format:

 ../OPUS Spectra/1_2019_11
 ../OPUS Spectra/2_2019_12
 ../OPUS Spectra/3_2020_01
 ...
 
 ../OPUS Spectra/10_2020_10