Solve Recurrence Relation with Lambert W-function

  • Context: Graduate 
  • Thread starter Thread starter spiceman
  • Start date Start date
  • Tags Tags
    Recurrence Relation
Click For Summary

Discussion Overview

The discussion revolves around solving a nonlinear recurrence relation of the form y[n+1] = y[n] + k * log(1 + (y[n]/k)), with a focus on transforming it into an explicit function. Participants explore various mathematical approaches and special cases related to the recurrence, including the potential application of the Lambert W-function.

Discussion Character

  • Exploratory
  • Mathematical reasoning
  • Debate/contested

Main Points Raised

  • One participant expresses difficulty in transforming the recurrence relation into an explicit function and suggests that the Lambert W-function might be relevant.
  • Another participant inquires about the range of the constant k, indicating its importance in the analysis.
  • Special cases for k are discussed, including the scenarios where k = 0 and k approaches infinity, leading to different behaviors of the recurrence relation.
  • A participant proposes a transformation of the recurrence relation using a new variable s[i], leading to a new equation involving s[i] and log(s[i]).
  • Another participant suggests a potential relationship for s[n] involving the logarithm of factorials, indicating a possible connection to combinatorial structures.
  • One participant provides numerical values generated from the recurrence relation with specific initial conditions and asks if others can identify a trend in the results.

Areas of Agreement / Disagreement

Participants have not reached a consensus on the solution to the recurrence relation. Multiple competing views and approaches remain, with ongoing exploration of different mathematical techniques and transformations.

Contextual Notes

Participants note that k is a constant greater than 0, typically around 0.5, which may influence the behavior of the recurrence relation. The discussion includes various assumptions and conditions that have not been fully resolved.

spiceman
Messages
3
Reaction score
0
Dear forum people,

for a nonlinear software I am writing I am having a hard time to transform a
recurrence relation to an explicit function. Maybe someone can help me along the right lines...

The recurrence relation is of the form (an exponential type function)

y[n+1] = y[n] + k * log(1 + (y[n]/k))

Now I know that simple linear recurrence functions like y[n+1] = y[n] + y[n] transform to explicit functions like y = 2^(n-1), but the tricky part is the log(1+...).
I have a feeling that the Lambert W-function could be a solution to the series
log(1+log(1+log(...))), but I am stuck at the moment. Even Mathematica fails on this problem with its function 'Rsolve'. Does anybody have an idea?
BTW, n is positive integer > 0 and y[n] > 0.

Greets and thanks in advance!
 
Mathematics news on Phys.org
What is the range for the constant k?
 
There are some special cases:

- if k = 0 you will have the limit of k * log (1+y(n)/k) = 0 therefore y(n+1) = y(n)
- if k -> infinity you will have k * log(1+y(n)/k)) = log((1+y(n)/k)^k) = log(e^y(n)) = y(n) and therefore y(n+1) = 2*y(n)
 
Let
1 + y/k = s{i]

so y = k * s - 1

Can we solve

k*s[i+1] - 1 = k*s - 1 + k log(s)

s[i+1] - s = log( s )
 
s[n] =\sum_{i=2}^{i=n} \log(s[i-1])

One thought is s[n] = \log((n-1)!)
 
I forgot to mention that k is a constant > 0.
Usually something like 0.5.

Greets.
 
Using y[0] = 1 and k = 1 the recurrence function will give
the following first 50 values:

1.000000 1.693147 2.683858 3.987818 5.594817 7.481101 9.618941
11.981580 14.545112 17.288858 20.195150 23.248922 26.437294 29.749198
33.175062 36.706558 40.336392 44.058135 47.866089 51.755172 55.720834
59.758976 63.865890 68.038212 72.272873 76.567063 80.918206 85.323927
89.782034 94.290495 98.847425 103.451069 108.099787 112.792050 117.526423 122.301559 127.116192 131.969130 136.859247 141.785480 146.746823 151.742323 156.771076 161.832221 166.924941 172.048458 177.202030 182.384948 187.596535 192.836145

Does anyone see a trend?
 

Similar threads

  • · Replies 12 ·
Replies
12
Views
3K
  • · Replies 11 ·
Replies
11
Views
2K
  • · Replies 3 ·
Replies
3
Views
2K
  • · Replies 3 ·
Replies
3
Views
2K
  • · Replies 13 ·
Replies
13
Views
2K
  • · Replies 12 ·
Replies
12
Views
2K
  • · Replies 8 ·
Replies
8
Views
4K
  • · Replies 2 ·
Replies
2
Views
2K
  • · Replies 1 ·
Replies
1
Views
2K
  • · Replies 18 ·
Replies
18
Views
3K