Related
I am trying to learn Julia for its potential use in parameter estimation. I am interested in estimating kinetic parameters of chemical reactions, which usually involves optimizing reaction parameters with multiple independent batches of experiments. I have successfully optimized a single batch, but need to expand the problem to use many different batches. In developing a sample problem, I am trying to optimize using two toy batches. I know there are probably smarter ways to do this (subject of a future question), but my current workflow involves calling an ODEProblem for each batch, calculating its loss against the data, and minimizing the sum of the residuals for the two batches. Unfortunately, I get an error when initiating the optimization with Optimization.jl. The current code and error are shown below:
using DifferentialEquations, Plots, DiffEqParamEstim
using Optimization, ForwardDiff, OptimizationOptimJL, OptimizationNLopt
using Ipopt, OptimizationGCMAES, Optimisers
using Random
#Experimental data, species B is NOT observed in the data
times = [0.0, 0.071875, 0.143750, 0.215625, 0.287500, 0.359375, 0.431250,
0.503125, 0.575000, 0.646875, 0.718750, 0.790625, 0.862500,
0.934375, 1.006250, 1.078125, 1.150000]
A_obs = [1.0, 0.552208, 0.300598, 0.196879, 0.101175, 0.065684, 0.045096,
0.028880, 0.018433, 0.011509, 0.006215, 0.004278, 0.002698,
0.001944, 0.001116, 0.000732, 0.000426]
C_obs = [0.0, 0.187768, 0.262406, 0.350412, 0.325110, 0.367181, 0.348264,
0.325085, 0.355673, 0.361805, 0.363117, 0.327266, 0.330211,
0.385798, 0.358132, 0.380497, 0.383051]
P_obs = [0.0, 0.117684, 0.175074, 0.236679, 0.234442, 0.270303, 0.272637,
0.274075, 0.278981, 0.297151, 0.297797, 0.298722, 0.326645,
0.303198, 0.277822, 0.284194, 0.301471]
#Create additional data sets for a multi data set optimization
#Simple noise added to data for testing
times_2 = times[2:end] .+ rand(range(-0.05,0.05,100))
P_obs_2 = P_obs[2:end] .+ rand(range(-0.05,0.05,100))
A_obs_2 = A_obs[2:end].+ rand(range(-0.05,0.05,100))
C_obs_2 = C_obs[2:end].+ rand(range(-0.05,0.05,100))
#ki = [2.78E+00, 1.00E-09, 1.97E-01, 3.04E+00, 2.15E+00, 5.27E-01] #Target optimized parameters
ki = [0.1, 0.1, 0.1, 0.1, 0.1, 0.1] #Initial guess of parameters
IC = [1.0, 0.0, 0.0, 0.0] #Initial condition for each species
tspan1 = (minimum(times),maximum(times)) #tuple timespan of data set 1
tspan2 = (minimum(times_2),maximum(times_2)) #tuple timespan of data set 2
# data = VectorOfArray([A_obs,C_obs,P_obs])'
data = vcat(A_obs',C_obs',P_obs') #Make multidimensional array containing all observed data for dataset1, transpose to match shape of ODEProblem output
data2 = vcat(A_obs_2',C_obs_2',P_obs_2') #Make multidimensional array containing all observed data for dataset2, transpose to match shape of ODEProblem output
#make dictionary containing data, time, and initial conditions
keys1 = ["A","B"]
keys2 = ["time","obs","IC"]
entryA =[times,data,IC]
entryB = [times_2, data2,IC]
nest=[Dict(zip(keys2,entryA)),Dict(zip(keys2,entryB))]
exp_dict = Dict(zip(keys1,nest)) #data dictionary
#rate equations in power law form r = k [A][B]
function rxn(x, k)
A = x[1]
B = x[2]
C = x[3]
P = x[4]
k1 = k[1]
k2 = k[2]
k3 = k[3]
k4 = k[4]
k5 = k[5]
k6 = k[6]
r1 = k1 * A
r2 = k2 * A * B
r3 = k3 * C * B
r4 = k4 * A
r5 = k5 * A
r6 = k6 * A * B
return [r1, r2, r3, r4, r5, r6] #returns reaction rate of each equation
end
#Mass balance differential equations
function mass_balances(di,x,args,t)
k = args
r = rxn(x, k)
di[1] = - r[1] - r[2] - r[4] - r[5] - r[6] #Species A
di[2] = + r[1] - r[2] - r[3] - r[6] #Species B
di[3] = + r[2] - r[3] + r[4] #Species C
di[4] = + r[3] + r[5] + r[6] #Species P
end
function ODESols(time,uo,parms)
time_init = (minimum(time),maximum(time))
prob = ODEProblem(mass_balances,uo,time_init,parms)
sol = solve(prob, Tsit5(), reltol=1e-8, abstol=1e-8,save_idxs = [1,3,4],saveat=time) #Integrate prob
return sol
end
function cost_function(data_dict,parms)
res_dict = Dict(zip(keys(data_dict),[0.0,0.0]))
for key in keys(data_dict)
pred = ODESols(data_dict[key]["time"],data_dict[key]["IC"],parms)
loss = L2Loss(data_dict[key]["time"],data_dict[key]["obs"])
err = loss(pred)
res_dict[key] = err
end
residual = sum(res_dict[key] for key in keys(res_dict))
#show typeof(residual)
return residual
end
lb = [0.0,0.0,0.0,0.0,0.0,0.0] #parameter lower bounds
ub = [10.0,10.0,10.0,10.0,10.0,10.0] #parameter upper bounds
optfun = Optimization.OptimizationFunction(cost_function,Optimization.AutoForwardDiff())
optprob = Optimization.OptimizationProblem(optfun,exp_dict, ki,lb=lb,ub=ub,reltol=1E-8) #Set up optimization problem
optsol=solve(optprob, BFGS(),maxiters=10000) #Solve optimization problem
println(optsol.u) #print solution
when I call optsol I get the error:
ERROR: MethodError: no method matching ForwardDiff.GradientConfig(::Optimization.var"#89#106"{OptimizationFunction{true, Optimization.AutoForwardDiff{nothing}, typeof(cost_function), Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, typeof(SciMLBase.DEFAULT_OBSERVED_NO_TIME), Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing}, Vector{Float64}}, ::Dict{String, Dict{String, Array{Float64}}}, ::ForwardDiff.Chunk{2})
Searching online suggests that the issue may be that my cost_function function is not generic enough for ForwardDiff to handle, however I am not sure how to identify where the issue is in this function, or whether it is related to the functions (mass_balances and rxn) that are called within cost_function. Another potential issue is that I am not calling the functions appropriately when building the OptimizationFunction or the OpptimizationProblem, but I cannot identify the issue here either.
Thank you for any suggestions and your help in troubleshooting this application!
res_dict = Dict(zip(keys(data_dict),[0.0,0.0]))
This dictionary is declared to the wrong type.
zerotype = zero(params[1])
res_dict = Dict(zip(keys(data_dict),[zerotype ,zerotype]))
or
res_dict = Dict(zip(keys(data_dict),zeros(eltype(params),2)))
Either way, you want your intermediate calculations to match the type of params when using AutoForwardDiff().
In addition to the variable type specification suggested by Chris, my model also had an issue with the order of the arguments of cost_function and how I passed the arguments to the problem in optprob. This solution was shown by Contradict here
I was trying to plot a logistic GAM model using the gratia package (since it uses ggplot2), however I would like the effects (or partial effects) to be plotted in terms of probabilities instead of log odds.
I have tried by hand using probabilities, however I prefer to use the gratia package. Would there be a way to plot the probabilities specifically using the package?
The model (I created some data):
set.seed(1)
Perf1 <- rlnorm(100)
Sex <- sample(c(rep(1, 40), rep(0, 60)))
set.seed(2)
Group <- sample(c(rep(1, 30), rep(0, 70)))
set.seed(3)
Perf2 <- rlnorm(200)
G <- sample(c(rep(1, 20), rep(0, 80)))
Age <- sample(c(rep(7, 15), rep(8, 20), rep(9, 30), rep(10, 10), rep(11, 15), rep(12, 10)))
sample_data <-data.frame(Age = Age,
Sex = Sex,
G = G,
Group = Group,
Perf1 = Perf1,
Perf2 = Perf1
)
gam_fit <- gam(Group ~ Age + Sex + G + s(Perf1, k = 20) +
s(Perf2, k = 20),
data = sample_data,
family = "binomial",
method="REML", select = F)
draw(gam_fit, parametric = T)
Plotting using gratia:
gam_fit
The effect or partial effect is on the log odds scale, while I would like probabilities instead, but I am unsure how to achieve this.
You'll have to add the model constant term and transform by the inverse of the link function:
draw(gam_fit, constant = coef(gam_fit)[1], fun = inv_link(gam_fit))
(and I'm not sure if constant or fun work on the parametric terms just now.)
I have a system given by this recursive relationship: xt = At xt-1 + bt. I wish to compute xt for all t, with At, bt and x0 given. Is there are built-in function for that? If I use a loop it would be extremely slow. Thanks!
There is sort of a way. Let's say you have your A matrices in a 3D tensor with shape (T, N, N), where T is the total number of time steps and N is the size of your vector. Similarly, B values are in a 2D tensor (T, N). The first step in the computation would be:
x1 = A[0] # x0 + B[0]
Where # represents matrix product. But you can convert this into a single matrix product. Suppose we add a value 1 at the end of x0, and we call that x0p (for prime):
x0p = tf.concat([x, [1]], axis=0)
And now we build a new 3D tensor Ap with shape (T, N+1, N+1), such that for each A[i] we concatenate B[i] as a new column, and then we add a row with N zeros and a single one at the end:
AwithB = tf.concat([tf.concat([A, tf.expand_dims(B, 2)], axis=2)], axis=1)
AnewRow = tf.concat([tf.zeros((T, 1, N), A.dtype), tf.ones((T, 1, 1), A.dtype)], axis=2)
Ap = tf.concat([AwithB, AnewRow], axis=1)
As it turns out, you can now say:
x1p = Ap[0] # x0p
And therefore:
x2p = Ap[1] # x1p = Ap[1] # Ap[0] # x0p
So we just need to compute all the matrix product of all matrices in Ap across the first dimension. Unfortunately, there does not seem to be a direct operation to compute that with TensorFlow, but you can do it relatively fast with tf.scan:
Ap_prod = tf.scan(tf.matmul, Ap)[-1]
And with that you just have to do:
xtp = Ap_prod # x0p
Here is a proof of concept (the code is tweaked to support single examples and batches, either in the A and B values or in the x)
import tensorflow as tf
def compute_state(a, b, x):
s = tf.shape(a)
t = s[-3]
n = s[-1]
# Add final 1 to x
xp = tf.concat([x, tf.ones_like(x[..., :1])], axis=-1)
# Add B column to A
a_b = tf.concat([tf.concat([a, tf.expand_dims(b, axis=-1)], axis=-1)], axis=-2)
# Make new final row for A
a_row = tf.concat([tf.zeros_like(a[..., :1, :]),
tf.ones_like(a[..., :1, :1])], axis=-1)
# Add new row to A
ap = tf.concat([a_b, a_row], axis=-2)
# Compute matrix product reduction
ap_prod = tf.scan(tf.matmul, ap)[..., -1, :, :]
# Compute final result
outp = tf.linalg.matvec(ap_prod, xp)
return outp[..., :-1]
#Test
tf.random.set_seed(0)
a = tf.random.uniform((10, 5, 5), -1, 1)
b = tf.random.uniform((10, 5), -1, 1)
x = tf.random.uniform((5,), -1, 1)
y = compute_state(a, b, x)
# Also works with batches of (a, b) or x
a = tf.random.uniform((100, 10, 5, 5), -1, 1)
b = tf.random.uniform((100, 10, 5), -1, 1)
x = tf.random.uniform((100, 5), -1, 1)
y = compute_state(a, b, x)
I've created a plot with geom_line and geom_ribbon (image 1) and the result is okay, but for the sake of aesthetics, I'd like the line and ribbon to be smoother. I know I can use geom_smooth for the line (image 2), but I'm not sure if it's possible to smooth the ribbon.I could create a geom_smooth line for the top and bottom lines of the ribbon (image 3), but is there anyway to fill in the space between those two lines?
A principled way to achieve what you want is to fit a GAM model to your data using the gam() function in mgcv and then apply the predict() function to that model over a finer grid of values for your predictor variable. The grid can cover the span defined by the range of observed values for your predictor variable. The R code below illustrates this process for a concrete example.
# load R packages
library(ggplot2)
library(mgcv)
# simulate some x and y data
# x = predictor; y = response
x <- seq(-10, 10, by = 1)
y <- 1 - 0.5*x - 2*x^2 + rnorm(length(x), mean = 0, sd = 20)
d <- data.frame(x,y)
# plot the simulated data
ggplot(data = d, aes(x,y)) +
geom_point(size=3)
# fit GAM model
m <- gam(y ~ s(x), data = d)
# define finer grid of predictor values
xnew <- seq(-10, 10, by = 0.1)
# apply predict() function to the fitted GAM model
# using the finer grid of x values
p <- predict(m, newdata = data.frame(x = xnew), se = TRUE)
str(p)
# plot the estimated mean values of y (fit) at given x values
# over the finer grid of x values;
# superimpose approximate 95% confidence band for the true
# mean values of y at given x values in the finer grid
g <- data.frame(x = xnew,
fit = p$fit,
lwr = p$fit - 1.96*p$se.fit,
upr = p$fit + 1.96*p$se.fit)
head(g)
theme_set(theme_bw())
ggplot(data = g, aes(x, fit)) +
geom_ribbon(aes(ymin = lwr, ymax = upr), fill = "lightblue") +
geom_line() +
geom_point(data = d, aes(x, y), shape = 1)
This same principle would apply if you were to fit a polynomial regression model to your data using the lm() function.
I'm trying to create a depth profile graph with the variables depth, distance and temperature. The data collected is from 9 different points with known distances between them (distance 5m apart, 9 stations, 9 different sets of data). The temperature readings are according to these 9 stations where a sonde was dropped directly down, taking readings of temperature every 2 seconds. Max depth at each of the 9 stations were taken from the boat also.
So the data I have is:
Depth at each of the 9 stations (y axis)
Temperature readings at each of the 9 stations, at around .2m intervals vertical until the bottom was reached (fill area)
distance between the stations, (x axis)
Is it possible to create a depth profile similar to this? (obviously without the greater resolution in this graph)
I've already tried messing around with ggplot2 and raster but I just can't seem to figure out how to do this.
One of the problems I've come across is how to make ggplot2 distinguish between say 5m depth temperature reading at station 1 and 5m temperature reading at station 5 since they have the same depth value.
Even if you can guide me towards another program that would allow me to create a graph like this, that would be great
[ REVISION ]
(Please comment me if you know more suitable interpolation methods, especially not needing to cut under bottoms data.)
ggplot() needs long data form.
library(ggplot2)
# example data
max.depths <- c(1.1, 4, 4.7, 7.7, 8.2, 7.8, 10.7, 12.1, 14.3)
depth.list <- sapply(max.depths, function(x) seq(0, x, 0.2))
temp.list <- list()
set.seed(1); for(i in 1:9) temp.list[[i]] <- sapply(depth.list[[i]], function(x) rnorm(1, 20 - x*0.5, 0.2))
set.seed(1); dist <- c(0, sapply(seq(5, 40, 5), function(x) rnorm(1, x, 1)))
dist.list <- sapply(1:9, function(x) rep(dist[x], length(depth.list[[x]])))
main.df <- data.frame(dist = unlist(dist.list), depth = unlist(depth.list) * -1, temp = unlist(temp.list))
# a raw graph
ggplot(main.df, aes(x = dist, y = depth, z = temp)) +
geom_point(aes(colour = temp), size = 1) +
scale_colour_gradientn(colours = topo.colors(10))
# a relatively raw graph (don't run with this example data)
ggplot(main.df, aes(x = dist, y = depth, z = temp)) +
geom_raster(aes(fill = temp)) + # geom_contour() +
scale_fill_gradientn(colours = topo.colors(10))
If you want a graph such like you showed, you have to do interpolation. Some packages give you spatial interpolation methods. In this example, I used akima package but you should think seriously that which interpolation methods to use.
I used nx = 300 and ny = 300 in below code but I think it would be better to decide those values carefully. Large nx and ny gives a high resolution graph, but don't foreget real nx and ny (in this example, real nx is only 9 and ny is 101).
library(akima); library(dplyr)
interp.data <- interp(main.df$dist, main.df$depth, main.df$temp, nx = 300, ny = 300)
interp.df <- interp.data %>% interp2xyz() %>% as.data.frame()
names(interp.df) <- c("dist", "depth", "temp")
# draw interp.df
ggplot(interp.df, aes(x = dist, y = depth, z = temp)) +
geom_raster(aes(fill = temp)) + # geom_contour() +
scale_fill_gradientn(colours = topo.colors(10))
# to think appropriateness of interpolation (raw and interpolation data)
ggplot(interp.df, aes(x = dist, y = depth, z = temp)) +
geom_raster(aes(fill = temp), alpha = 0.3) + # interpolation
scale_fill_gradientn(colours = topo.colors(10)) +
geom_point(data = main.df, aes(colour = temp), size = 1) + # raw
scale_colour_gradientn(colours = topo.colors(10))
Bottoms don't match !!I found ?interp says "interpolation only within convex hull!", oops... I'm worrid about the interpolation around the problem-area, is it OK ? If no problem, you need only cut the data under the bottoms. If not, ... I can't answer immediately (below is an example code to cut).
bottoms <- max.depths * -1
# calculate bottom values using linear interpolation
approx.bottoms <- approx(dist, bottoms, n = 300) # n must be the same value as interp()'s nx
# change temp values under bottom into NA
library(dplyr)
interp.cut.df <- interp.df %>% cbind(bottoms = approx.bottoms$y) %>%
mutate(temp = ifelse(depth >= bottoms, temp, NA)) %>% select(-bottoms)
ggplot(interp.cut.df, aes(x = dist, y = depth, z = temp)) +
geom_raster(aes(fill = temp)) +
scale_fill_gradientn(colours = topo.colors(10)) +
geom_point(data = main.df, size = 1)
If you want to use stat_contour
It is harder to use stat_contour than geom_raster because it needs a regular grid form. As far as I see your graph, your data (depth and distance) don't form a regular grid, it means it is much difficult to use stat_contour with your raw data. So I used interp.cut.df to draw a contour plot. And stat_contour have a endemic problem (see How to fill in the contour fully using stat_contour), so you need to expand your data.
library(dplyr)
# 1st: change NA into a temp's out range value (I used 0)
interp.contour.df <- interp.cut.df
interp.contour.df[is.na(interp.contour.df)] <- 0
# 2nd: expand the df (It's a little complex, so please use this function)
contour.support.func <- function(df) {
colname <- names(df)
names(df) <- c("x", "y", "z")
Range <- as.data.frame(sapply(df, range))
Dim <- as.data.frame(t(sapply(df, function(x) length(unique(x)))))
arb_z = Range$z[1] - diff(Range$z)/20
df2 <- rbind(df,
expand.grid(x = c(Range$x[1] - diff(Range$x)/20, Range$x[2] + diff(Range$x)/20),
y = seq(Range$y[1], Range$y[2], length = Dim$y), z = arb_z),
expand.grid(x = seq(Range$x[1], Range$x[2], length = Dim$x),
y = c(Range$y[1] - diff(Range$y)/20, Range$y[2] + diff(Range$y)/20), z = arb_z))
names(df2) <- colname
return(df2)
}
interp.contour.df2 <- contour.support.func(interp.contour.df)
# 3rd: check the temp range (these values are used to define contour's border (breaks))
range(interp.cut.df$temp, na.rm=T) # 12.51622 20.18904
# 4th: draw ... the bottom border is dirty !!
ggplot(interp.contour.df2, aes(x = dist, y = depth, z = temp)) +
stat_contour(geom="polygon", breaks = seq(12.51622, 20.18904, length = 11), aes(fill = ..level..)) +
coord_cartesian(xlim = range(dist), ylim = range(bottoms), expand = F) + # cut expanded area
scale_fill_gradientn(colours = topo.colors(10)) # breaks's length is 11, so 10 colors are needed
# [Note]
# You can define the contour's border values (breaks) and colors.
contour.breaks <- c(12.5, 13.5, 14.5, 15.5, 16.5, 17.5, 18.5, 19.5, 20.5)
# = seq(12.5, 20.5, 1) or seq(12.5, 20.5, length = 9)
contour.colors <- c("darkblue", "cyan3", "cyan1", "green3", "green", "yellow2","pink", "darkred")
# breaks's length is 9, so 8 colors are needed.
# 5th: vanish the bottom border by bottom line
approx.df <- data.frame(dist = approx.bottoms$x, depth = approx.bottoms$y, temp = 0) # 0 is dummy value
ggplot(interp.contour.df2, aes(x = dist, y = depth, z = temp)) +
stat_contour(geom="polygon", breaks = contour.breaks, aes(fill = ..level..)) +
coord_cartesian(xlim=range(dist), ylim=range(bottoms), expand = F) +
scale_fill_gradientn(colours = contour.colors) +
geom_line(data = approx.df, lwd=1.5, color="gray50")
bonus: legend technic
library(dplyr)
interp.contour.df3 <- interp.contour.df2 %>% mutate(temp2 = cut(temp, breaks = contour.breaks))
interp.contour.df3$temp2 <- factor(interp.contour.df3$temp2, levels = rev(levels(interp.contour.df3$temp2)))
ggplot(interp.contour.df3, aes(x = dist, y = depth, z = temp)) +
stat_contour(geom="polygon", breaks = contour.breaks, aes(fill = ..level..)) +
coord_cartesian(xlim=range(dist), ylim=range(bottoms), expand = F) +
scale_fill_gradientn(colours = contour.colors, guide = F) + # add guide = F
geom_line(data = approx.df, lwd=1.5, color="gray50") +
geom_point(aes(colour = temp2), pch = 15, alpha = 0) + # add
guides(colour = guide_legend(override.aes = list(colour = rev(contour.colors), alpha = 1, cex = 5))) + # add
labs(colour = "temp") # add
You want to treat this as a 3-D surface with temperature as the z dimension. The given plot is a contour plot and it looks like ggplot2 can do that with stat_contour.
I'm not sure how the contour lines are computed (often it's linear interpolation along a Delaunay triangulation). If you want more control over how to interpolate between your x/y grid points, you can calculate a surface model first and feed those z coordinates into ggplot2.