Add weights to the data, bound at a maximum.

Description

Add weights to the data, bound at a maximum.

Usage

Bounded_weights(S, weight.upper.bound = 30)

Arguments

Value

A vector with upper bounds for weights

Examples

Bounded_weights(Sm, weight.upper.bound = 30)

Cluster things

Description

Cluster things

Usage

Cluster(
  Data,
  minSamplesPerCluster,
  row_ids = NULL,
  dist_method = "euclidean",
  hclust_method = "ward.D2"
)

Arguments

Value

A named list of length two. The first element "cluster.list" is a list of clusters, and the second element "cluster.plot" the cluster analysis object (dendogram) that can be plotted.

Examples

Cluster.result <- Cluster(Sm, 14)
Cluster.result$cluster.list
plot(Cluster.result$cluster.plot)

Calculate the mean condition number for randomized F matrices

Description

Performs multiple simulations with randomized F matrices within given bounds to assess the numerical stability of the system.

Usage

Condition_test(S, Fn, min.val = NULL, max.val = NULL)

Arguments

Value

Numeric value representing the mean condition number from 1000 simulations

Examples

# Create sample matrices
Fmat <- as.matrix(phytoclass::Fm)
S <- as.matrix(phytoclass::Sm)
min_max <- phytoclass::min_max
min.val <- min_max[[3]]
max.val <- min_max[[4]]
# Use only non-chla columns for Fmat and S, as in simulated_annealing
Fmat_sub <- Fmat[, -ncol(Fmat)]
S_sub <- S[, -ncol(S)]
# Calculate mean condition number
cond <- phytoclass:::Condition_test(S_sub, Fmat_sub, min.val, max.val)

Conduit between minimise_elements function and Fac_F_R of steepest descent algorithm.

Description

Conduit between minimise_elements function and Fac_F_R of steepest descent algorithm.

Usage

Conduit(Fmat, place, S, cm, c_num = c(1, 2, 3))

Arguments

Value

A list containing three elements: 1: The modified F matrix 2: Number of iterations or modifications made 3: The original F matrix before modifications

Examples

Fmat <- as.matrix(phytoclass::Fm)
S <- as.matrix(phytoclass::Sm)
S_weights <- as.numeric(phytoclass:::Bounded_weights(S))
place <- which(Fmat[, seq(ncol(Fmat) - 2)] > 0)

# Get F_initial from NNLS_MF as done in Steepest_Descent
F_initial <- phytoclass::NNLS_MF(Fmat, S, S_weights)

result <- phytoclass:::Conduit(F_initial[[1]], place, S, S_weights, c_num = 3)

Sets the default minimum and maximum values for phytoplankton groups pigment ratios. To use this function, pigment and phytoplankton group names will need to fit the naming criteria of phytoclass.

Description

Sets the default minimum and maximum values for phytoplankton groups pigment ratios. To use this function, pigment and phytoplankton group names will need to fit the naming criteria of phytoclass.

Usage

Default_min_max(min_max, Fmat)

Arguments

Value

A list containing two elements: 1: Vector of minimum values for each non-zero pigment ratio 2: Vector of maximum values for each non-zero pigment ratio

Examples

# Create a sample F matrix
Fmat <- phytoclass::Fm

# Create min_max data frame
min_max <- phytoclass::min_max
result <- phytoclass:::Default_min_max(min_max, Fmat)

Part of the steepest descent algorithm that works to reduce error given the S and F matrices.

Description

Part of the steepest descent algorithm that works to reduce error given the S and F matrices.

Usage

Fac_F_RR(Fmat, vary, S, cm, fac_rr = c(1, 2, 3), place = NULL)

Arguments

Value

A list containing two elements: 1: Updated F matrix after optimization 2: Vector of indices

Examples

 # Setup based on Minimise_elements_comb usage
 Fmat <- as.matrix(phytoclass::Fm)
 S <- as.matrix(phytoclass::Sm)
 cm <- as.numeric(phytoclass:::Bounded_weights(S))
 place <- which(Fmat[, seq(ncol(Fmat) - 2)] > 0)

 # Get F.new from Conduit as done in Minimise_elements_comb
 f <- phytoclass:::Conduit(Fmat, place, S, cm, c_num = 3)
 F.new <- f[[1]]

 # Set place1 as done in Minimise_elements_comb
 place1 <- place # place1 = place when c1_num != 1

 # Run Fac_F_RR
 result <- phytoclass:::Fac_F_RR(F.new, vary = place, place = place1, S, cm, fac_rr = 3)

Fm data

Description

Fm data

Usage

Fm

Format

Fm

A data frame with 9 rows and 15 columns:

chl_c1

XX

Per

XX

X19but

XX

...

Source

XX

Fp data

Description

Fp data

Usage

Fp

Format

Fp

A data frame with 9 rows and 15 columns:

chl_c1

XX

Per

XX

X19but

XX

...

Source

XX

Remove any column values that average 0. Further to this, also remove phytoplankton groups from the F matrix if their diagnostic pigment isn’t present.

Description

Remove any column values that average 0. Further to this, also remove phytoplankton groups from the F matrix if their diagnostic pigment isn’t present.

Usage

Matrix_checks(S, Fmat)

Arguments

Value

Named list with new S and Fmat matrices

Examples

MC <- Matrix_checks(Sm, Fm)  
Snew <- MC$Snew

Part of the steepest descent algorithm

Description

Part of the steepest descent algorithm

Usage

Minimise_elements_comb(Fmat, place, S, cm, c1_num = c(1, 2, 3))

Arguments

Value

A list containing the F matrix with updated ratios, the RMSE of new estimates, and an updated C matrix of estimated group contribution.

Examples

 Fmat <- as.matrix(phytoclass::Fm)
 S <- as.matrix(phytoclass::Sm)
 S_weights <- as.numeric(phytoclass:::Bounded_weights(S))
 place <- which(Fmat[, seq(ncol(Fmat) - 2)] > 0)

 # Get F_initial from NNLS_MF as done in Steepest_Descent
 F_initial <- phytoclass::NNLS_MF(Fmat, S, S_weights)

 # Run Minimise_elements_comb with c1_num = 3 (as in Steepest_Descent)
 result <- phytoclass:::Minimise_elements_comb(
   F_initial[[1]], place, S, S_weights, c1_num = 3
 )

Performs the non-negative matrix factorisation for given phytoplankton pigments and pigment ratios, to attain an estimate of phytoplankton class abundances.

Description

Performs the non-negative matrix factorisation for given phytoplankton pigments and pigment ratios, to attain an estimate of phytoplankton class abundances.

Usage

NNLS_MF(Fn, S, S_weights = NULL)

Arguments

Value

A list containing

1. The F matrix (pigment: Chl a) ratios

2. The root mean square error (RMSE)

3. The C matrix (class abundances for each group)

Examples

 Fmat <- as.matrix(phytoclass::Fm)
 S <- as.matrix(phytoclass::Sm)
 S_weights <- as.numeric(phytoclass:::Bounded_weights(S))
 place <- which(Fmat[, seq(ncol(Fmat) - 2)] > 0)
 num.loops <- 2
 # Run Steepest_Descent
 result <- phytoclass:::Steepest_Descent(Fmat, place, S, S_weights, num.loops)

Perform matrix factorisation for phytoplankton pigments and pigments ratios

Description

Performs the non-negative matrix factorisation for given phytoplankton pigments and pigment ratios, to attain an estimate of phytoplankton class abundances.

Usage

NNLS_MF_Final(Fn, S, S_Chl, S_weights, S_dvChl = NULL)

Arguments

Details

Unlike NNLS_ML(), it also removes any weighting and normalisation, and also multiplies relative abundances by chlorophyll values to determine the biomass of phytoplankton groups.

Value

A list containing the following elements:

Examples

 Fmat <- as.matrix(phytoclass::Fm)
 S <- as.matrix(phytoclass::Sm)
 S_weights <- as.numeric(phytoclass:::Bounded_weights(S))
 S_Chl <- S[, ncol(S)]
 # Run NNLS_MF_Final
 result <- phytoclass:::NNLS_MF_Final(Fmat, S, S_Chl, S_weights)

This function normalises each column in F to row sum

Description

This function normalises each column in F to row sum

Usage

Normalise_F(Fmat)

Arguments

Value

A list consisting of two components: - a matrix of pigment ratios normalized to row sums - a vector of row sums

Normalise matrix to row sum

Description

This function normalises each column in S to row sum

Usage

Normalise_S(S)

Arguments

Value

A matrix

Examples

# Create a sample matrix
S <- as.matrix(phytoclass::Sm)
normalized <- phytoclass:::Normalise_S(S)

Normalize F matrix specifically for Prochlorococcus pigments

Description

Normalizes pigment ratios differently for Prochlorococcus vs other groups, using divinyl chlorophyll a for Prochlorococcus and chlorophyll a for others.

Usage

Prochloro_Normalise_F(Fmat)

Arguments

Value

A list containing: 1: Normalized matrix where each row is scaled by its biomass marker 2: Vector of row sums from the scaled matrix

Examples

# Create sample F matrix with Prochlorococcus
Fmat <- as.matrix(phytoclass::Fp)
result <- phytoclass:::Prochloro_Normalise_F(Fmat)

Selects a random neighbour for a subset of non-zero pigments that are outside the min and max bounds for the simulated annealing algorithm, specifically handling Prochlorococcus pigments.

Description

Selects a random neighbour for a subset of non-zero pigments that are outside the min and max bounds for the simulated annealing algorithm, specifically handling Prochlorococcus pigments.

Usage

Prochloro_Random_Neighbour(
  f_new,
  Temp,
  chlv,
  chlvp,
  N,
  place,
  S,
  S_weights,
  minF,
  maxF
)

Arguments

Value

A list containing:

Examples

 Fmat <- as.matrix(phytoclass::Fp)
 S <- as.matrix(phytoclass::Sp)
 S_weights <- as.numeric(phytoclass:::Bounded_weights(S))
 place <- which(Fmat[, seq(ncol(Fmat) - 2)] > 0)

 # Get min_max from package data
 min_max_mat <- phytoclass:::Default_min_max(phytoclass::min_max, Fmat[, -ncol(Fmat)])

 # Get bounds using Prochloro_Wrangling as in simulated_annealing_Prochloro
 f_c <- Fmat
 W0 <- phytoclass:::Prochloro_Wrangling(f_c, min_max_mat[[1]], min_max_mat[[2]])
 minF <- W0[[1]]
 maxF <- W0[[2]]

 # Extract chlv and chlvp from the matrix
 chlv <- f_c[, ncol(f_c)] # Chl a (Tchla)
 chlvp <- f_c[, ncol(f_c) - 1] # dvChl a

 Temp <- 0.5
 N <- place

 # Run Prochloro_Random_Neighbour
 result <- phytoclass:::Prochloro_Random_Neighbour(
   f_c, Temp, chlv, chlvp, N, place, S, S_weights, minF, maxF
 )

Prochloro Wrangling

Description

Prochloro Wrangling

Usage

Prochloro_Wrangling(Fl, min.val, max.val)

Arguments

Value

A list containing vectorized min/max bounds and current values for Prochlorococcus-aware normalization

Select a random neighbour when the previous random neighbour is beyond the minimum or maximum value. Part of the simulated annealing algorithm.

Description

Select a random neighbour when the previous random neighbour is beyond the minimum or maximum value. Part of the simulated annealing algorithm.

Usage

Random_neighbour(f_new, Temp, chlv, N, place, S, S_weights, minF, maxF)

Arguments

Value

A list containing:

Examples

 # Setup based on simulated_annealing usage
 Fmat <- as.matrix(phytoclass::Fm)
 S <- as.matrix(phytoclass::Sm)
 S_weights <- as.numeric(phytoclass:::Bounded_weights(S))
 place <- which(Fmat[, seq(ncol(Fmat) - 2)] > 0)
 min_max <- phytoclass::min_max
 minF <- min_max[[3]][seq_along(place)]
 maxF <- min_max[[4]][seq_along(place)]
 chlv <- rep(1, nrow(Fmat)) # typical usage in simulated_annealing
 Temp <- 0.5
 N <- place
 # Run Random_neighbour
 result <- phytoclass:::Random_neighbour(Fmat, Temp, chlv, N, place, S, S_weights, minF, maxF)

Randomise value by applying scaling factors within specified bounds. Small values (< 0.001) are set to 0.001 to avoid numerical issues.

Description

Randomise value by applying scaling factors within specified bounds. Small values (< 0.001) are set to 0.001 to avoid numerical issues.

Usage

Randomise_elements(x, min.scaler, max.scaler)

Arguments

Value

A numeric value between xmin.scaler and xmax.scaler, rounded to 4 decimals

Examples

# Randomize a single value
x <- 0.5
new_value <- phytoclass:::Randomise_elements(x, 0.99, 1.01)  # +/- 1% change

# Handle small values
small_x <- 0.0005
new_small <- phytoclass:::Randomise_elements(small_x, 0.99, 1.01)  # will use 0.001

Select the new F matrix element with lowest error in the steepest descent algorithm. Randomly modifies a single element and checks if the modification reduces the error.

Description

Select the new F matrix element with lowest error in the steepest descent algorithm. Randomly modifies a single element and checks if the modification reduces the error.

Usage

Replace_Rand(Fmat, i, S, cm, min.scaler, max.scaler)

Arguments

Value

A list containing:

Examples

 # Setup based on Fac_F_RR usage
 Fmat <- as.matrix(phytoclass::Fm)
 S <- as.matrix(phytoclass::Sm)
 cm <- as.numeric(phytoclass:::Bounded_weights(S))

 # Get Fmat as a list from NNLS_MF (as used in Fac_F_RR)
 Fmat_list <- phytoclass::NNLS_MF(Fmat, S, cm)

 # Test with a single index
 i <- 1 # first non-zero element to modify
 min.scaler <- 0.99
 max.scaler <- 1.01

 # Run Replace_Rand
 result <- phytoclass:::Replace_Rand(Fmat_list, i, S, cm, min.scaler, max.scaler)

Apply the steepest descent algorithm to optimize pigment ratios in phytoplankton classification. Loosely wraps Seepest_Descent function.

Description

Apply the steepest descent algorithm to optimize pigment ratios in phytoplankton classification. Loosely wraps Seepest_Descent function.

Usage

SAALS(Ft, min.value, max.value, place, S, cm, num.loops)

Arguments

Value

A list containing: 1: The optimized F matrix 2: Final RMSE value

Examples

 Fmat <- as.matrix(phytoclass::Fm)
 S <- as.matrix(phytoclass::Sm)
 S_weights <- as.numeric(phytoclass:::Bounded_weights(S))
 place <- which(Fmat[, seq(ncol(Fmat) - 2)] > 0)
 num.loops <- 2
 # Run SAALS
 result <- phytoclass:::SAALS(Fmat, NULL, NULL, place, S, S_weights, num.loops)

Sm data

Description

Sm data

Usage

Sm

Format

Sm

A data frame with 29 rows and 15 columns:

chl_c1

XX

Per

XX

X19but

XX

...

Source

XX

Sp data

Description

Sp data

Usage

Sp

Format

Sp

A data frame with 29 rows and 15 columns:

chl_c1

XX

Per

XX

X19but

XX

...

Source

XX

Stand-alone version of steepest descent algorithm. This is similar to the CHEMTAX steepest descent algorithm. It is not required to use this function, and as results are not bound by minimum and maximum, results may be unrealistic.

Description

Stand-alone version of steepest descent algorithm. This is similar to the CHEMTAX steepest descent algorithm. It is not required to use this function, and as results are not bound by minimum and maximum, results may be unrealistic.

Usage

Steepest_Desc(Fmat, S, num.loops)

Arguments

Value

A list containing

1. The F matrix (pigment: Chl a) ratios

2. RMSE (Root Mean Square Error)

3. Condition number

4. class abundances

5. Figure (plot of results)

6. MAE (Mean Absolute Error)

Examples

MC <- Matrix_checks(Sm,Fm)
Snew <- MC$Snew
Fnew <- MC$Fnew
SDRes <- Steepest_Desc(Fnew,Snew, num.loops = 20)

Performs the steepest descent algorithm for a set number of iterations to optimize the F matrix of pigment ratios.

Description

Performs the steepest descent algorithm for a set number of iterations to optimize the F matrix of pigment ratios.

Usage

Steepest_Descent(Fmat, place, S, S_weights, num.loops)

Arguments

Value

A list containing: 1: The optimized F matrix 2: Final RMSE value 3: The C matrix (class abundances for each group)

Examples

 Fmat <- as.matrix(phytoclass::Fm)
 S <- as.matrix(phytoclass::Sm)
 S_weights <- as.numeric(phytoclass:::Bounded_weights(S))
 place <- which(Fmat[, seq(ncol(Fmat) - 2)] > 0)
 num.loops <- 2
 # Run Steepest_Descent
 result <- phytoclass:::Steepest_Descent(Fmat, place, S, S_weights, num.loops)

Apply weights to F/S matrices by diagonal multiplication

Description

Apply weights to F/S matrices by diagonal multiplication

Usage

Weight_error(S, cm)

Arguments

Value

A matrix with weighted columns (S %*% diag(cm))

Examples

# Create sample matrix and weights
S <- as.matrix(phytoclass::Sm)
Fmat <- as.matrix(phytoclass::Fm)
S_weights <- as.numeric(phytoclass:::Bounded_weights(S))
weighted <- phytoclass:::Weight_error(Fmat, S_weights)

Wrangle data to vectors

Description

Converts data-types and selects data for randomisation in the simulated annealing algorithm

Usage

Wrangling(Fl, min.val, max.val)

Arguments

Value

A list containing following components:
- A vector Fmin with the minimum pigment ratio values
- A vector Fmax with the maximum pigment ratio values
- A vector SE with the current pigment ratio values
- A vector chlv with the pigment ratio values for the last column in Fl

Convergence Figure

Description

A figure to show the pigment ratios for each phytoplankton group for each iteration.

Usage

convergence_figure(fm_iter, niter = NULL)

Arguments

Value

A figure with each pigment ratio per iteration per group

Examples

# ADD_EXAMPLES_HERE

min_max data

Description

min_max data

Usage

min_max

Format

min_max

A data frame with 51 rows and 4 columns:

class

XX

Pig_Abbrev

XX

min

XX

max

max

...

Source

XX

Phytoplankton Class Abundance Figure

Description

This function plots the class abundances as output by simulated_annealing.

Usage

phyto_figure(c_matrix)

Arguments

Value

A stacked line plot with sample number on x axis, chl a concentrations on y axis, and phytoplankton groups as colors

Examples

# ADD_EXAMPLES_HERE

This is the main phytoclass algorithm. It performs simulated annealing algorithm for S and F matrices. See the examples (Fm, Sm) for how to set up matrices, and the vignette for more detailed instructions. Different pigments and phytoplankton groups may be used.

Description

This is the main phytoclass algorithm. It performs simulated annealing algorithm for S and F matrices. See the examples (Fm, Sm) for how to set up matrices, and the vignette for more detailed instructions. Different pigments and phytoplankton groups may be used.

Usage

simulated_annealing(
  S,
  Fmat = NULL,
  user_defined_min_max = NULL,
  do_matrix_checks = TRUE,
  niter = 500,
  step = 0.009,
  weight.upper.bound = 30,
  verbose = TRUE,
  seed = NULL,
  check_converge = 100,
  alt_pro_name = NULL
)

Arguments

Value

A list containing

1. Fmat matrix

2. RMSE (Root Mean Square Error)

3. condition number

4. Class abundances

5. Figure (plot of results)

6. MAE (Mean Absolute Error)

7. Error

8. F_mat_iter

9. converge_plot

Examples

# Using the built-in matrices Sm and Fm
set.seed(5326)
sa.example <- simulated_annealing(Sm, Fm, niter = 5)
sa.example$Figure

Perform simulated annealing algorithm for samples with divinyl chlorophyll and prochlorococcus. Chlorophyll must be the final column of both S and F matrices, with Divinyl Chlorophyll a the 2nd to last column. See how the example Sp and Fp matrices are organised.

Description

Perform simulated annealing algorithm for samples with divinyl chlorophyll and prochlorococcus. Chlorophyll must be the final column of both S and F matrices, with Divinyl Chlorophyll a the 2nd to last column. See how the example Sp and Fp matrices are organised.

Usage

simulated_annealing_Prochloro(
  S,
  Fmat = NULL,
  user_defined_min_max = NULL,
  do_matrix_checks = TRUE,
  niter = 500,
  step = 0.009,
  weight.upper.bound = 30,
  verbose = TRUE,
  seed = NULL,
  check_converge = 100
)

Arguments

Value

A list containing

1. Fmat matrix

2. RMSE (Root Mean Square Error)

3. condition number

4. Class abundances

5. Figure (plot of results)

6. MAE (Mean Absolute Error)

7. Error

Examples

# Using the built-in matrices Sp and Fp.
set.seed(5326)
sa.example <- simulated_annealing_Prochloro(Sp, Fp, niter = 1)
sa.example$Figure

Vectorise a matrix and keep non-zero elements

Description

Turn each non-zero element of the F-matrix into a vector

Usage

vectorise(Fmat)

Arguments

Value

A vector of non-zero pigment elements