For from a range(0, value goal) - for-loop

'Hi, I'm trying to calculate how many months to achieve a million dollars with compounding interest and monthly investments. There are my tries.'
'This first code work, but I want to replace the 92 in the rage with a compare formula like fv >= 1000000.'
'When I place the range like here, it doesn't work.'

Try while-loop may help:
pv = 130000 # present value
i = 4000 # regular monthly investment
r = 0.1375 # annual interest rate
n = 0 # number of months
# for n in range(0, 92):
fv = 0
while fv < 1000000:
fv = pv * (1 + r / 12) ** n + i * (((1 + r / 12) ** n - 1) / (r / 12))
n += 1 #don't forget
print(fv)
print(n)
You need to manually increase the value of n

Related

Ruby algorithms loops codewars

I got stuck with below task and spent about 3 hours trying to figure it out.
Task description: A man has a rather old car being worth $2000. He saw a secondhand car being worth $8000. He wants to keep his old car until he can buy the secondhand one.
He thinks he can save $1000 each month but the prices of his old car and of the new one decrease of 1.5 percent per month. Furthermore this percent of loss increases by 0.5 percent at the end of every two months. Our man finds it difficult to make all these calculations.
How many months will it take him to save up enough money to buy the car he wants, and how much money will he have left over?
My code so far:
def nbMonths(startPriceOld, startPriceNew, savingperMonth, percentLossByMonth)
dep_value_old = startPriceOld
mth_count = 0
total_savings = 0
dep_value_new = startPriceNew
mth_count_new = 0
while startPriceOld != startPriceNew do
if startPriceOld >= startPriceNew
return mth_count = 0, startPriceOld - startPriceNew
end
dep_value_new = dep_value_new - (dep_value_new * percentLossByMonth / 100)
mth_count_new += 1
if mth_count_new % 2 == 0
dep_value_new = dep_value_new - (dep_value_new * 0.5) / 100
end
dep_value_old = dep_value_old - (dep_value_old * percentLossByMonth / 100)
mth_count += 1
total_savings += savingperMonth
if mth_count % 2 == 0
dep_value_old = dep_value_old - (dep_value_old * 0.5) / 100
end
affordability = total_savings + dep_value_old
if affordability >= dep_value_new
return mth_count, affordability - dep_value_new
end
end
end
print nbMonths(2000, 8000, 1000, 1.5) # Expected result[6, 766])
The data are as follows.
op = 2000.0 # current old car value
np = 8000.0 # current new car price
sv = 1000.0 # annual savings
dr = 0.015 # annual depreciation, both cars (1.5%)
cr = 0.005. # additional depreciation every two years, both cars (0.5%)
After n >= 0 months the man's (let's call him "Rufus") savings plus the value of his car equal
sv*n + op*(1 - n*dr - (cr + 2*cr + 3*cr +...+ (n/2)*cr))
where n/2 is integer division. As
cr + 2*cr + 3*cr +...+ (n/2)*cr = cr*((1+2+..+n)/2) = cr*(1+n/2)*(n/2)
the expression becomes
sv*n + op*(1 - n*dr - cr*(1+(n/2))*(n/2))
Similarly, after n years the cost of the car he wants to purchase will fall to
np * (1 - n*dr - cr*(1+(n/2))*(n/2))
If we set these two expressions equal we obtain the following.
sv*n + op - op*dr*n - op*cr*(n/2) - op*cr*(n/2)**2 =
np - np*dr*n - np*cr*(n/2) - np*cr*(n/2)**2
which reduces to
cr*(np-op)*(n/2)**2 + (sv + dr*(np-op))*n + cr*(np-op)*(n/2) - (np-op) = 0
or
cr*(n/2)**2 + (sv/(np-op) + dr)*n + cr*(n/2) - 1 = 0
If we momentarily treat (n/2) as a float division, this expression reduces to a quadratic.
(cr/4)*n**2 + (sv/(np-op) + dr + cr/2)*n - 1 = 0
= a*n**2 + b*n + c = 0
where
a = cr/4 = 0.005/4 = 0.00125
b = sv/(np-op) + dr + cr/(2*a) = 1000.0/(8000-2000) + 0.015 + 0.005/2 = 0.18417
c = -1
Incidentally, Rufus doesn't have a computer, but he does have an HP 12c calculator his grandfather gave him when he was a kid, which is perfectly adequate for these simple calculations.
The roots are computed as follows.
(-b + Math.sqrt(b**2 - 4*a*c))/(2*a) #=> 5.24
(-b - Math.sqrt(b**2 - 4*a*c))/(2*a) #=> -152.58
It appears that Rufus can purchase the new vehicle (if it's still for sale) in six years. Had we been able able to solve the above equation for n/2 using integer division it might have turned out that Rufus would have had to wait longer. That’s because for a given n both cars would have depreciated less (or at least not not more), and because the car to be purchased is more expensive than the current car, the difference in values would be greater than that obtained with the float approximation for 1/n. We need to check that, however. After n years, Rufus' savings and the value of his beater will equal
sv*n + op*(1 - dr*n - cr*(1+(n/2))*(n/2))
= 1000*n + 2000*(1 - 0.015*n - 0.005*(1+(n/2))*(n/2))
For n = 6 this equals
1000*6 + 2000*(1 - 0.015*6 - 0.005*(1+(6/2))*(6/2))
= 1000*6 + 2000*(1 - 0.015*6 - 0.005*(1+3)*3)
= 1000*6 + 2000*0.85
= 7700
The cost of Rufus' dream car after n years will be
np * (1 - dr*n - cr*(1+(n/2))*(n/2))
= 8000 * (1 - 0.015*n - 0.005*(1+(n/2))*(n/2))
For n=6 this becomes
8000 * (1 - 0.015*6 - 0.005*(1+(6/2))*(6/2))
= 8000*0.85
= 6800
(Notice that the factor 0.85 is the same in both calculations.)
Yes, Rufus will be able to buy the car in 6 years.
def nbMonths(old, new, savings, percent)
percent = percent.fdiv(100)
current_savings = 0
months = 0
loop do
break if current_savings + old >= new
current_savings += savings
old -= old * percent
new -= new * percent
months += 1
percent += 0.005 if months.odd?
end
[months, (current_savings + old - new).round]
end

Filling a matrix using parallel processing in Julia

I'm trying to speed up the solution time for a dynamic programming problem in Julia (v. 0.5.0), via parallel processing. The problem involves choosing the optimal values for every element of a 1073 x 19 matrix at every iteration, until successive matrix differences fall within a tolerance. I thought that, within each iteration, filling in the values for each element of the matrix could be parallelized. However, I'm seeing a huge performance degradation using SharedArray, and I'm wondering if there's a better way to approach parallel processing for this problem.
I construct the arguments for the function below:
est_params = [.788,.288,.0034,.1519,.1615,.0041,.0077,.2,0.005,.7196]
r = 0.015
tau = 0.35
rho =est_params[1]
sigma =est_params[2]
delta = 0.15
gamma =est_params[3]
a_capital =est_params[4]
lambda1 =est_params[5]
lambda2 =est_params[6]
s =est_params[7]
theta =est_params[8]
mu =est_params[9]
p_bar_k_ss =est_params[10]
beta = (1+r)^(-1)
sigma_range = 4
gz = 19
gp = 29
gk = 37
lnz=collect(linspace(-sigma_range*sigma,sigma_range*sigma,gz))
z=exp(lnz)
gk_m = fld(gk,2)
# Need to add mu somewhere to k_ss
k_ss = (theta*(1-tau)/(r+delta))^(1/(1-theta))
k=cat(1,map(i->k_ss*((1-delta)^i),collect(1:gk_m)),map(i->k_ss/((1-delta)^i),collect(1:gk_m)))
insert!(k,gk_m+1,k_ss)
sort!(k)
p_bar=p_bar_k_ss*k_ss
p = collect(linspace(-p_bar/2,p_bar,gp))
#Tauchen
N = length(z)
Z = zeros(N,1)
Zprob = zeros(Float32,N,N)
Z[N] = lnz[length(z)]
Z[1] = lnz[1]
zstep = (Z[N] - Z[1]) / (N - 1)
for i=2:(N-1)
Z[i] = Z[1] + zstep * (i - 1)
end
for a = 1 : N
for b = 1 : N
if b == 1
Zprob[a,b] = 0.5*erfc(-((Z[1] - mu - rho * Z[a] + zstep / 2) / sigma)/sqrt(2))
elseif b == N
Zprob[a,b] = 1 - 0.5*erfc(-((Z[N] - mu - rho * Z[a] - zstep / 2) / sigma)/sqrt(2))
else
Zprob[a,b] = 0.5*erfc(-((Z[b] - mu - rho * Z[a] + zstep / 2) / sigma)/sqrt(2)) -
0.5*erfc(-((Z[b] - mu - rho * Z[a] - zstep / 2) / sigma)/sqrt(2))
end
end
end
# Collecting tauchen results in a 2 element array of linspace and array; [2] gets array
# Zprob=collect(tauchen(gz, rho, sigma, mu, sigma_range))[2]
Zcumprob=zeros(Float32,gz,gz)
# 2 in cumsum! denotes the 2nd dimension, i.e. columns
cumsum!(Zcumprob, Zprob,2)
gm = gk * gp
control=zeros(gm,2)
for i=1:gk
control[(1+gp*(i-1)):(gp*i),1]=fill(k[i],(gp,1))
control[(1+gp*(i-1)):(gp*i),2]=p
end
endog=copy(control)
E=Array(Float32,gm,gm,gz)
for h=1:gm
for m=1:gm
for j=1:gz
# set the nonzero net debt indicator
if endog[h,2]<0
p_ind=1
else
p_ind=0
end
# set the investment indicator
if (control[m,1]-(1-delta)*endog[h,1])!=0
i_ind=1
else
i_ind=0
end
E[m,h,j] = (1-tau)*z[j]*(endog[h,1]^theta) + control[m,2]-endog[h,2]*(1+r*(1-tau)) +
delta*endog[h,1]*tau-(control[m,1]-(1-delta)*endog[h,1]) -
(i_ind*gamma*endog[h,1]+endog[h,1]*(a_capital/2)*(((control[m,1]-(1-delta)*endog[h,1])/endog[h,1])^2)) +
s*endog[h,2]*p_ind
elem = E[m,h,j]
if E[m,h,j]<0
E[m,h,j]=elem+lambda1*elem-.5*lambda2*elem^2
else
E[m,h,j]=elem
end
end
end
end
I then constructed the function with serial processing. The two for loops iterate through each element to find the largest value in a 1072-sized (=the gm scalar argument in the function) array:
function dynam_serial(E,gm,gz,beta,Zprob)
v = Array(Float32,gm,gz )
fill!(v,E[cld(gm,2),cld(gm,2),cld(gz,2)])
Tv = Array(Float32,gm,gz)
# Set parameters for the loop
convcrit = 0.0001 # chosen convergence criterion
diff = 1 # arbitrary initial value greater than convcrit
while diff>convcrit
exp_v=v*Zprob'
for h=1:gm
for j=1:gz
Tv[h,j]=findmax(E[:,h,j] + beta*exp_v[:,j])[1]
end
end
diff = maxabs(Tv - v)
v=copy(Tv)
end
end
Timing this, I get:
#time dynam_serial(E,gm,gz,beta,Zprob)
> 106.880008 seconds (91.70 M allocations: 203.233 GB, 15.22% gc time)
Now, I try using Shared Arrays to benefit from parallel processing. Note that I reconfigured the iteration so that I only have one for loop, rather than two. I also use v=deepcopy(Tv); otherwise, v is copied as an Array object, rather than a SharedArray:
function dynam_parallel(E,gm,gz,beta,Zprob)
v = SharedArray(Float32,(gm,gz),init = S -> S[Base.localindexes(S)] = myid() )
fill!(v,E[cld(gm,2),cld(gm,2),cld(gz,2)])
# Set parameters for the loop
convcrit = 0.0001 # chosen convergence criterion
diff = 1 # arbitrary initial value greater than convcrit
while diff>convcrit
exp_v=v*Zprob'
Tv = SharedArray(Float32,gm,gz,init = S -> S[Base.localindexes(S)] = myid() )
#sync #parallel for hj=1:(gm*gz)
j=cld(hj,gm)
h=mod(hj,gm)
if h==0;h=gm;end;
#async Tv[h,j]=findmax(E[:,h,j] + beta*exp_v[:,j])[1]
end
diff = maxabs(Tv - v)
v=deepcopy(Tv)
end
end
Timing the parallel version; and using a 4-core 2.5 GHz I7 processor with 16GB of memory, I get:
addprocs(3)
#time dynam_parallel(E,gm,gz,beta,Zprob)
> 164.237208 seconds (2.64 M allocations: 201.812 MB, 0.04% gc time)
Am I doing something incorrect here? Or is there a better way to approach parallel processing in Julia for this particular problem? I've considered using Distributed Arrays, but it's difficult for me to see how to apply them to the present problem.
UPDATE:
Per #DanGetz and his helpful comments, I turned instead to trying to speed up the serial processing version. I was able to get performance down to 53.469780 seconds (67.36 M allocations: 103.419 GiB, 19.12% gc time) through:
1) Upgrading to 0.6.0 (saved about 25 seconds), which includes the helpful #views macro.
2) Preallocating the main array I'm trying to fill in (Tv), per the section on Preallocating Outputs in the Julia Performance Tips: https://docs.julialang.org/en/latest/manual/performance-tips/. (saved another 25 or so seconds)
The biggest remaining slow-down seems to be coming from the add_vecs function, which sums together subarrays of two larger matrices. I've tried devectorizing and using BLAS functions, but haven't been able to produce better performance.
In any event, the improved code for dynam_serial is below:
function add_vecs(r::Array{Float32},h::Int,j::Int,E::Array{Float32},exp_v::Array{Float32},beta::Float32)
#views r=E[:,h,j] + beta*exp_v[:,j]
return r
end
function dynam_serial(E::Array{Float32},gm::Int,gz::Int,beta::Float32,Zprob::Array{Float32})
v = Array{Float32}(gm,gz)
fill!(v,E[cld(gm,2),cld(gm,2),cld(gz,2)])
Tv = Array{Float32}(gm,gz)
r = Array{Float32}(gm)
# Set parameters for the loop
convcrit = 0.0001 # chosen convergence criterion
diff = 1 # arbitrary initial value greater than convcrit
while diff>convcrit
exp_v=v*Zprob'
for h=1:gm
for j=1:gz
#views Tv[h,j]=findmax(add_vecs(r,h,j,E,exp_v,beta))[1]
end
end
diff = maximum(abs,Tv - v)
v=copy(Tv)
end
return Tv
end
If add_vecs seems to be the critical function, writing an explicit for loop could offer more optimization. How does the following benchmark:
function add_vecs!(r::Array{Float32},h::Int,j::Int,E::Array{Float32},
exp_v::Array{Float32},beta::Float32)
#inbounds for i=1:size(E,1)
r[i]=E[i,h,j] + beta*exp_v[i,j]
end
return r
end
UPDATE
To continue optimizing dynam_serial I have tried to remove more allocations. The result is:
function add_vecs_and_max!(gm::Int,r::Array{Float64},h::Int,j::Int,E::Array{Float64},
exp_v::Array{Float64},beta::Float64)
#inbounds for i=1:gm
r[i] = E[i,h,j]+beta*exp_v[i,j]
end
return findmax(r)[1]
end
function dynam_serial(E::Array{Float64},gm::Int,gz::Int,
beta::Float64,Zprob::Array{Float64})
v = Array{Float64}(gm,gz)
fill!(v,E[cld(gm,2),cld(gm,2),cld(gz,2)])
r = Array{Float64}(gm)
exp_v = Array{Float64}(gm,gz)
# Set parameters for the loop
convcrit = 0.0001 # chosen convergence criterion
diff = 1.0 # arbitrary initial value greater than convcrit
while diff>convcrit
A_mul_Bt!(exp_v,v,Zprob)
diff = -Inf
for h=1:gm
for j=1:gz
oldv = v[h,j]
newv = add_vecs_and_max!(gm,r,h,j,E,exp_v,beta)
v[h,j]= newv
diff = max(diff, oldv-newv, newv-oldv)
end
end
end
return v
end
Switching the functions to use Float64 should increase speed (as CPUs are inherently optimized for 64-bit word lengths). Also, using the mutating A_mul_Bt! directly saves another allocation. Avoiding the copy(...) by switching the arrays v and Tv.
How do these optimizations improve your running time?
2nd UPDATE
Updated the code in the UPDATE section to use findmax. Also, changed dynam_serial to use v without Tv, as there was no need to save the old version except for the diff calculation, which is now done inside the loop.
Here's the code I copied-and-pasted, provided by Dan Getz above. I include the array and scalar definitions exactly as I ran them. Performance was: 39.507005 seconds (11 allocations: 486.891 KiB) when running #time dynam_serial(E,gm,gz,beta,Zprob).
using SpecialFunctions
est_params = [.788,.288,.0034,.1519,.1615,.0041,.0077,.2,0.005,.7196]
r = 0.015
tau = 0.35
rho =est_params[1]
sigma =est_params[2]
delta = 0.15
gamma =est_params[3]
a_capital =est_params[4]
lambda1 =est_params[5]
lambda2 =est_params[6]
s =est_params[7]
theta =est_params[8]
mu =est_params[9]
p_bar_k_ss =est_params[10]
beta = (1+r)^(-1)
sigma_range = 4
gz = 19 #15 #19
gp = 29 #19 #29
gk = 37 #25 #37
lnz=collect(linspace(-sigma_range*sigma,sigma_range*sigma,gz))
z=exp.(lnz)
gk_m = fld(gk,2)
# Need to add mu somewhere to k_ss
k_ss = (theta*(1-tau)/(r+delta))^(1/(1-theta))
k=cat(1,map(i->k_ss*((1-delta)^i),collect(1:gk_m)),map(i->k_ss/((1-delta)^i),collect(1:gk_m)))
insert!(k,gk_m+1,k_ss)
sort!(k)
p_bar=p_bar_k_ss*k_ss
p = collect(linspace(-p_bar/2,p_bar,gp))
#Tauchen
N = length(z)
Z = zeros(N,1)
Zprob = zeros(Float64,N,N)
Z[N] = lnz[length(z)]
Z[1] = lnz[1]
zstep = (Z[N] - Z[1]) / (N - 1)
for i=2:(N-1)
Z[i] = Z[1] + zstep * (i - 1)
end
for a = 1 : N
for b = 1 : N
if b == 1
Zprob[a,b] = 0.5*erfc(-((Z[1] - mu - rho * Z[a] + zstep / 2) / sigma)/sqrt(2))
elseif b == N
Zprob[a,b] = 1 - 0.5*erfc(-((Z[N] - mu - rho * Z[a] - zstep / 2) / sigma)/sqrt(2))
else
Zprob[a,b] = 0.5*erfc(-((Z[b] - mu - rho * Z[a] + zstep / 2) / sigma)/sqrt(2)) -
0.5*erfc(-((Z[b] - mu - rho * Z[a] - zstep / 2) / sigma)/sqrt(2))
end
end
end
# Collecting tauchen results in a 2 element array of linspace and array; [2] gets array
# Zprob=collect(tauchen(gz, rho, sigma, mu, sigma_range))[2]
Zcumprob=zeros(Float64,gz,gz)
# 2 in cumsum! denotes the 2nd dimension, i.e. columns
cumsum!(Zcumprob, Zprob,2)
gm = gk * gp
control=zeros(gm,2)
for i=1:gk
control[(1+gp*(i-1)):(gp*i),1]=fill(k[i],(gp,1))
control[(1+gp*(i-1)):(gp*i),2]=p
end
endog=copy(control)
E=Array(Float64,gm,gm,gz)
for h=1:gm
for m=1:gm
for j=1:gz
# set the nonzero net debt indicator
if endog[h,2]<0
p_ind=1
else
p_ind=0
end
# set the investment indicator
if (control[m,1]-(1-delta)*endog[h,1])!=0
i_ind=1
else
i_ind=0
end
E[m,h,j] = (1-tau)*z[j]*(endog[h,1]^theta) + control[m,2]-endog[h,2]*(1+r*(1-tau)) +
delta*endog[h,1]*tau-(control[m,1]-(1-delta)*endog[h,1]) -
(i_ind*gamma*endog[h,1]+endog[h,1]*(a_capital/2)*(((control[m,1]-(1-delta)*endog[h,1])/endog[h,1])^2)) +
s*endog[h,2]*p_ind
elem = E[m,h,j]
if E[m,h,j]<0
E[m,h,j]=elem+lambda1*elem-.5*lambda2*elem^2
else
E[m,h,j]=elem
end
end
end
end
function add_vecs_and_max!(gm::Int,r::Array{Float64},h::Int,j::Int,E::Array{Float64},
exp_v::Array{Float64},beta::Float64)
maxr = -Inf
#inbounds for i=1:gm r[i] = E[i,h,j]+beta*exp_v[i,j]
maxr = max(r[i],maxr)
end
return maxr
end
function dynam_serial(E::Array{Float64},gm::Int,gz::Int,
beta::Float64,Zprob::Array{Float64})
v = Array{Float64}(gm,gz)
fill!(v,E[cld(gm,2),cld(gm,2),cld(gz,2)])
Tv = Array{Float64}(gm,gz)
r = Array{Float64}(gm)
exp_v = Array{Float64}(gm,gz)
# Set parameters for the loop
convcrit = 0.0001 # chosen convergence criterion
diff = 1.0 # arbitrary initial value greater than convcrit
while diff>convcrit
A_mul_Bt!(exp_v,v,Zprob)
diff = -Inf
for h=1:gm
for j=1:gz
Tv[h,j]=add_vecs_and_max!(gm,r,h,j,E,exp_v,beta)
diff = max(abs(Tv[h,j]-v[h,j]),diff)
end
end
(v,Tv)=(Tv,v)
end
return v
end
Now, here's another version of the algorithm and inputs. The functions are similar to what Dan Getz suggested, except that I use findmax rather than an iterated max function to find the array maximum. In the input construction, I am using both Float32 and mixing different bit-types together. However, I've consistently achieved better performance this way: 24.905569 seconds (1.81 k allocations: 46.829 MiB, 0.01% gc time). But it's not clear at all why.
using SpecialFunctions
est_params = [.788,.288,.0034,.1519,.1615,.0041,.0077,.2,0.005,.7196]
r = 0.015
tau = 0.35
rho =est_params[1]
sigma =est_params[2]
delta = 0.15
gamma =est_params[3]
a_capital =est_params[4]
lambda1 =est_params[5]
lambda2 =est_params[6]
s =est_params[7]
theta =est_params[8]
mu =est_params[9]
p_bar_k_ss =est_params[10]
beta = Float32((1+r)^(-1))
sigma_range = 4
gz = 19
gp = 29
gk = 37
lnz=collect(linspace(-sigma_range*sigma,sigma_range*sigma,gz))
z=exp(lnz)
gk_m = fld(gk,2)
# Need to add mu somewhere to k_ss
k_ss = (theta*(1-tau)/(r+delta))^(1/(1-theta))
k=cat(1,map(i->k_ss*((1-delta)^i),collect(1:gk_m)),map(i->k_ss/((1-delta)^i),collect(1:gk_m)))
insert!(k,gk_m+1,k_ss)
sort!(k)
p_bar=p_bar_k_ss*k_ss
p = collect(linspace(-p_bar/2,p_bar,gp))
#Tauchen
N = length(z)
Z = zeros(N,1)
Zprob = zeros(Float32,N,N)
Z[N] = lnz[length(z)]
Z[1] = lnz[1]
zstep = (Z[N] - Z[1]) / (N - 1)
for i=2:(N-1)
Z[i] = Z[1] + zstep * (i - 1)
end
for a = 1 : N
for b = 1 : N
if b == 1
Zprob[a,b] = 0.5*erfc(-((Z[1] - mu - rho * Z[a] + zstep / 2) / sigma)/sqrt(2))
elseif b == N
Zprob[a,b] = 1 - 0.5*erfc(-((Z[N] - mu - rho * Z[a] - zstep / 2) / sigma)/sqrt(2))
else
Zprob[a,b] = 0.5*erfc(-((Z[b] - mu - rho * Z[a] + zstep / 2) / sigma)/sqrt(2)) -
0.5*erfc(-((Z[b] - mu - rho * Z[a] - zstep / 2) / sigma)/sqrt(2))
end
end
end
# Collecting tauchen results in a 2 element array of linspace and array; [2] gets array
# Zprob=collect(tauchen(gz, rho, sigma, mu, sigma_range))[2]
Zcumprob=zeros(Float32,gz,gz)
# 2 in cumsum! denotes the 2nd dimension, i.e. columns
cumsum!(Zcumprob, Zprob,2)
gm = gk * gp
control=zeros(gm,2)
for i=1:gk
control[(1+gp*(i-1)):(gp*i),1]=fill(k[i],(gp,1))
control[(1+gp*(i-1)):(gp*i),2]=p
end
endog=copy(control)
E=Array(Float32,gm,gm,gz)
for h=1:gm
for m=1:gm
for j=1:gz
# set the nonzero net debt indicator
if endog[h,2]<0
p_ind=1
else
p_ind=0
end
# set the investment indicator
if (control[m,1]-(1-delta)*endog[h,1])!=0
i_ind=1
else
i_ind=0
end
E[m,h,j] = (1-tau)*z[j]*(endog[h,1]^theta) + control[m,2]-endog[h,2]*(1+r*(1-tau)) +
delta*endog[h,1]*tau-(control[m,1]-(1-delta)*endog[h,1]) -
(i_ind*gamma*endog[h,1]+endog[h,1]*(a_capital/2)*(((control[m,1]-(1-delta)*endog[h,1])/endog[h,1])^2)) +
s*endog[h,2]*p_ind
elem = E[m,h,j]
if E[m,h,j]<0
E[m,h,j]=elem+lambda1*elem-.5*lambda2*elem^2
else
E[m,h,j]=elem
end
end
end
end
function add_vecs!(gm::Int,r::Array{Float32},h::Int,j::Int,E::Array{Float32},
exp_v::Array{Float32},beta::Float32)
#inbounds #views for i=1:gm
r[i]=E[i,h,j] + beta*exp_v[i,j]
end
return r
end
function dynam_serial(E::Array{Float32},gm::Int,gz::Int,beta::Float32,Zprob::Array{Float32})
v = Array{Float32}(gm,gz)
fill!(v,E[cld(gm,2),cld(gm,2),cld(gz,2)])
Tv = Array{Float32}(gm,gz)
# Set parameters for the loop
convcrit = 0.0001 # chosen convergence criterion
diff = 1.00000 # arbitrary initial value greater than convcrit
iter=0
exp_v=Array{Float32}(gm,gz)
r=Array{Float32}(gm)
while diff>convcrit
A_mul_Bt!(exp_v,v,Zprob)
for h=1:gm
for j=1:gz
Tv[h,j]=findmax(add_vecs!(gm,r,h,j,E,exp_v,beta))[1]
end
end
diff = maximum(abs,Tv - v)
(v,Tv)=(Tv,v)
end
return v
end

split rectangle in cells / random coordinates / store in array in FORTRAN

I would like to split a rectangle in cells. In each cell it should be create a random coordinate (y, z).
The wide and height of the rectangle are known (initialW / initalH).
The size of the cells are calculated (dy / dz).
The numbers, in how many cells the rectangle to be part, are known. (numberCellsY / numberCellsZ)
Here my Code in Fortran to split the rectangle in Cells:
yRVEMin = 0.0
yRVEMax = initialW
dy = ( yRVEMax - yRVEMin ) / numberCellsY
zRVEMin = 0.0
zRVEMax = initialH
dz = ( zRVEMax - zRVEMin ) / numberCellsZ
do i = 1, numberCellsY
yMin(i) = (i-1)*dy
yMax(i) = i*dy
end do
do j = 1, numberCellsZ
zMin(j) = (j-1)*dz
zMax(j) = j*dz
end do
Now I would like to produce a random coordinate in each cell. The problem for me is, to store the coodinates in an array. It does not necessarily all be stored in one array, but as least as possible.
To fill the cells with coordinates it should start at the bottom left cell, go through the rows (y-direction), and after the last cell (numberCellsY) jump a column higher (z-dicrection) and start again by the first cell of the new row at left side. That should be made so long until a prescribed number (nfibers) is reached.
Here a deplorable try to do it:
call random_seed
l = 0
do k = 1 , nfibers
if (l < numberCellsY) then
l = l + 1
else
l = 1
end if
call random_number(y)
fiberCoordY(k) = yMin(l) + y * (yMax(l) - yMin(l))
end do
n = 0
do m = 1 , nfibers
if (n < numberCellsZ) then
n = n + 1
else
n = 1
end if
call random_number(z)
fiberCoordZ(m) = zMin(n) + z * (zMax(n) - zMin(n))
end do
The output is not what I want! fiberCoordZ should be stay on (zMin(1) / zMax(1) as long as numberCellsY-steps are reached.
The output for following settings:
nfibers = 9
numberCellsY = 3
numberCellsZ = 3
initialW = 9.0
initialH = 9.0
My random output for fiberCoordY is:
1.768946 3.362770 8.667685 1.898700 5.796713 8.770239 2.463412 3.546694 7.074708
and for fiberCoordZ is:
2.234807 5.213032 6.762228 2.948657 5.937295 8.649946 0.6795220 4.340364 8.352566
In this case the first 3 numbers of fiberCoordz should have a value between 0.0 and 3.0. Than number 4 - 6 a value between 3.0 and 6.0. And number 7 - 9 a value bewtween 6.0 - 9.0.
How can I solve this? If somebody has a solution with a better approach, please post it!
Thanks
Looking at
n = 0
do m = 1 , nfibers
if (n < numberCellsZ) then
n = n + 1
else
n = 1
end if
call random_number(z)
fiberCoordZ(m) = zMin(n) + z * (zMax(n) - zMin(n))
end do
we see that the z coordinate offset (the bottom cell boundary of interest) is being incremented inappropriately: for each consecutive nfibers/numberCellsZ coordinates n should be constant.
n should be incremented only every numberCellsY iterations, so perhaps a condition like
if (MOD(m, numberCellsY).eq.1) n=n+1
would be better.
Thanks francescalus! It works fine.
I added a little more for the case that nfibers > numberCellsY*numberCellsZ
n=0
do m = 1 , nfibers
if (MOD(m, numberCellsY).eq.1 .and. (n < numberCellsY)) then
n=n+1
end if
if (MOD(m, numberCellsY*numberCellsZ).eq.1 ) then
n = 1
end if
call random_number(z)
fiberCoordZ(m) = zMin(n) + z * (zMax(n) - zMin(n))
end do

Non-linear counter

So I have a counter. It is supposed to calculate the current amount of something. To calculate this, I know the start date, and start amount, and the amount to increment the counter by each second. Easy peasy. The tricky part is that the growth is not quite linear. Every day, the increment amount increases by a set amount. I need to recreate this algorithmically - basically figure out the exact value at the current date based on the starting value, the amount incremented over time, and the amount the increment has increased over time.
My target language is Javascript, but pseudocode is fine too.
Based on AB's solution:
var now = new Date();
var startDate1 = new Date("January 1 2010");
var days1 = (now - startDate1) / 1000 / 60 / 60 / 24;
var startNumber1 = 9344747520;
var startIncrement1 = 463;
var dailyIncrementAdjustment1 = .506;
var currentIncrement = startIncrement1 + (dailyIncrementAdjustment1 * days1);
startNumber1 = startNumber1 + (days1 / 2) * (2 * startIncrement1 + (days1 - 1) * dailyIncrementAdjustment1);
Does that look reasonable to you guys?
It's a quadratic function. If t is the time passed, then it's the usual at2+bt+c, and you can figure out a,b,c by substituting the results for the first 3 seconds.
Or: use the formula for the arithmetic progression sum, where a1 is the initial increment, and d is the "set amount" you refer to. Just don't forget to add your "start amount" to what the formula gives you.
If x0 is the initial amount, d is the initial increment, and e is the "set amount" to increase the incerement, it comes to
x0 + (t/2)*(2d + (t-1)*e)
If I understand your question correctly, you have an initial value x_0, an initial increment per second of d_0 and an increment adjustment of e per day. That is, on day one the increment per second is d_0, on day two the increment per second is d_0 + e, etc.
Then, we note that the increment per second at time t is
d(t) = d_0 + floor(t / S) * e
where S is the number of seconds per day and t is the number of seconds that have elapsed since t = t_0. Then
x = x_0 + sum_{k < floor(t / S)} S * d(k) + S * (t / S - floor(t / S)) * d(t)
is the formula that you are seeking. From here, you can simplify this to
x = x_0 + S * floor(t / S) d_0 + S * e * (floor(t / S) - 1) * floor(t / S) / 2.
use strict; use warnings;
my $start = 0;
my $stop = 100;
my $current = $start;
for my $day ( 1 .. 100 ) {
$current += ($day / 10);
last unless $current < $stop;
printf "Day: %d\tLeft %.2f\n", $day, (1 - $current/$stop);
}
Output:
Day: 1 Left 1.00
Day: 2 Left 1.00
Day: 3 Left 0.99
Day: 4 Left 0.99
Day: 5 Left 0.98
...
Day: 42 Left 0.10
Day: 43 Left 0.05
Day: 44 Left 0.01

Tickmark algorithm for a graph axis

I'm looking for an algorithm that places tick marks on an axis, given a range to display, a width to display it in, and a function to measure a string width for a tick mark.
For example, given that I need to display between 1e-6 and 5e-6 and a width to display in pixels, the algorithm would determine that I should put tickmarks (for example) at 1e-6, 2e-6, 3e-6, 4e-6, and 5e-6. Given a smaller width, it might decide that the optimal placement is only at the even positions, i.e. 2e-6 and 4e-6 (since putting more tickmarks would cause them to overlap).
A smart algorithm would give preference to tickmarks at multiples of 10, 5, and 2. Also, a smart algorithm would be symmetric around zero.
As I didn't like any of the solutions I've found so far, I implemented my own. It's in C# but it can be easily translated into any other language.
It basically chooses from a list of possible steps the smallest one that displays all values, without leaving any value exactly in the edge, lets you easily select which possible steps you want to use (without having to edit ugly if-else if blocks), and supports any range of values. I used a C# Tuple to return three values just for a quick and simple demonstration.
private static Tuple<decimal, decimal, decimal> GetScaleDetails(decimal min, decimal max)
{
// Minimal increment to avoid round extreme values to be on the edge of the chart
decimal epsilon = (max - min) / 1e6m;
max += epsilon;
min -= epsilon;
decimal range = max - min;
// Target number of values to be displayed on the Y axis (it may be less)
int stepCount = 20;
// First approximation
decimal roughStep = range / (stepCount - 1);
// Set best step for the range
decimal[] goodNormalizedSteps = { 1, 1.5m, 2, 2.5m, 5, 7.5m, 10 }; // keep the 10 at the end
// Or use these if you prefer: { 1, 2, 5, 10 };
// Normalize rough step to find the normalized one that fits best
decimal stepPower = (decimal)Math.Pow(10, -Math.Floor(Math.Log10((double)Math.Abs(roughStep))));
var normalizedStep = roughStep * stepPower;
var goodNormalizedStep = goodNormalizedSteps.First(n => n >= normalizedStep);
decimal step = goodNormalizedStep / stepPower;
// Determine the scale limits based on the chosen step.
decimal scaleMax = Math.Ceiling(max / step) * step;
decimal scaleMin = Math.Floor(min / step) * step;
return new Tuple<decimal, decimal, decimal>(scaleMin, scaleMax, step);
}
static void Main()
{
// Dummy code to show a usage example.
var minimumValue = data.Min();
var maximumValue = data.Max();
var results = GetScaleDetails(minimumValue, maximumValue);
chart.YAxis.MinValue = results.Item1;
chart.YAxis.MaxValue = results.Item2;
chart.YAxis.Step = results.Item3;
}
Take the longest of the segments about zero (or the whole graph, if zero is not in the range) - for example, if you have something on the range [-5, 1], take [-5,0].
Figure out approximately how long this segment will be, in ticks. This is just dividing the length by the width of a tick. So suppose the method says that we can put 11 ticks in from -5 to 0. This is our upper bound. For the shorter side, we'll just mirror the result on the longer side.
Now try to put in as many (up to 11) ticks in, such that the marker for each tick in the form i*10*10^n, i*5*10^n, i*2*10^n, where n is an integer, and i is the index of the tick. Now it's an optimization problem - we want to maximize the number of ticks we can put in, while at the same time minimizing the distance between the last tick and the end of the result. So assign a score for getting as many ticks as we can, less than our upper bound, and assign a score to getting the last tick close to n - you'll have to experiment here.
In the above example, try n = 1. We get 1 tick (at i=0). n = 2 gives us 1 tick, and we're further from the lower bound, so we know that we have to go the other way. n = 0 gives us 6 ticks, at each integer point point. n = -1 gives us 12 ticks (0, -0.5, ..., -5.0). n = -2 gives us 24 ticks, and so on. The scoring algorithm will give them each a score - higher means a better method.
Do this again for the i * 5 * 10^n, and i*2*10^n, and take the one with the best score.
(as an example scoring algorithm, say that the score is the distance to the last tick times the maximum number of ticks minus the number needed. This will likely be bad, but it'll serve as a decent starting point).
Funnily enough, just over a week ago I came here looking for an answer to the same question, but went away again and decided to come up with my own algorithm. I am here to share, in case it is of any use.
I wrote the code in Python to try and bust out a solution as quickly as possible, but it can easily be ported to any other language.
The function below calculates the appropriate interval (which I have allowed to be either 10**n, 2*10**n, 4*10**n or 5*10**n) for a given range of data, and then calculates the locations at which to place the ticks (based on which numbers within the range are divisble by the interval). I have not used the modulo % operator, since it does not work properly with floating-point numbers due to floating-point arithmetic rounding errors.
Code:
import math
def get_tick_positions(data: list):
if len(data) == 0:
return []
retpoints = []
data_range = max(data) - min(data)
lower_bound = min(data) - data_range/10
upper_bound = max(data) + data_range/10
view_range = upper_bound - lower_bound
num = lower_bound
n = math.floor(math.log10(view_range) - 1)
interval = 10**n
num_ticks = 1
while num <= upper_bound:
num += interval
num_ticks += 1
if num_ticks > 10:
if interval == 10 ** n:
interval = 2 * 10 ** n
elif interval == 2 * 10 ** n:
interval = 4 * 10 ** n
elif interval == 4 * 10 ** n:
interval = 5 * 10 ** n
else:
n += 1
interval = 10 ** n
num = lower_bound
num_ticks = 1
if view_range >= 10:
copy_interval = interval
else:
if interval == 10 ** n:
copy_interval = 1
elif interval == 2 * 10 ** n:
copy_interval = 2
elif interval == 4 * 10 ** n:
copy_interval = 4
else:
copy_interval = 5
first_val = 0
prev_val = 0
times = 0
temp_log = math.log10(interval)
if math.isclose(lower_bound, 0):
first_val = 0
elif lower_bound < 0:
if upper_bound < -2*interval:
if n < 0:
copy_ub = round(upper_bound*10**(abs(temp_log) + 1))
times = copy_ub // round(interval*10**(abs(temp_log) + 1)) + 2
else:
times = upper_bound // round(interval) + 2
while first_val >= lower_bound:
prev_val = first_val
first_val = times * copy_interval
if n < 0:
first_val *= (10**n)
times -= 1
first_val = prev_val
times += 3
else:
if lower_bound > 2*interval:
if n < 0:
copy_ub = round(lower_bound*10**(abs(temp_log) + 1))
times = copy_ub // round(interval*10**(abs(temp_log) + 1)) - 2
else:
times = lower_bound // round(interval) - 2
while first_val < lower_bound:
first_val = times*copy_interval
if n < 0:
first_val *= (10**n)
times += 1
if n < 0:
retpoints.append(first_val)
else:
retpoints.append(round(first_val))
val = first_val
times = 1
while val <= upper_bound:
val = first_val + times * interval
if n < 0:
retpoints.append(val)
else:
retpoints.append(round(val))
times += 1
retpoints.pop()
return retpoints
When passing in the following three data-points to the function
points = [-0.00493, -0.0003892, -0.00003292]
... the output I get (as a list) is as follows:
[-0.005, -0.004, -0.003, -0.002, -0.001, 0.0]
When passing this:
points = [1.399, 38.23823, 8309.33, 112990.12]
... I get:
[0, 20000, 40000, 60000, 80000, 100000, 120000]
When passing this:
points = [-54, -32, -19, -17, -13, -11, -8, -4, 12, 15, 68]
... I get:
[-60, -40, -20, 0, 20, 40, 60, 80]
... which all seem to be a decent choice of positions for placing ticks.
The function is written to allow 5-10 ticks, but that could easily be changed if you so please.
Whether the list of data supplied contains ordered or unordered data it does not matter, since it is only the minimum and maximum data points within the list that matter.
This simple algorithm yields an interval that is multiple of 1, 2, or 5 times a power of 10. And the axis range gets divided in at least 5 intervals. The code sample is in java language:
protected double calculateInterval(double range) {
double x = Math.pow(10.0, Math.floor(Math.log10(range)));
if (range / x >= 5)
return x;
else if (range / (x / 2.0) >= 5)
return x / 2.0;
else
return x / 5.0;
}
This is an alternative, for minimum 10 intervals:
protected double calculateInterval(double range) {
double x = Math.pow(10.0, Math.floor(Math.log10(range)));
if (range / (x / 2.0) >= 10)
return x / 2.0;
else if (range / (x / 5.0) >= 10)
return x / 5.0;
else
return x / 10.0;
}
I've been using the jQuery flot graph library. It's open source and does axis/tick generation quite well. I'd suggest looking at it's code and pinching some ideas from there.

Resources