Ruby Program for finding local maxima doesn't pass one test - arrays

Question

I have a program for finding peaks (local maxima) in ruby that passes all but one test. Personally, I think my program is ok, but maybe I'm not taking into account an assumed neighborhood size that isn't specified in the problem, but was suggested by someone else who also attempted it.
Here's what I have so far.
def pick_peaks(arr)
pos = []
peaks =[]
peak_set = {pos: [], peaks: []}
for i in 1..arr.length-2
if arr[i-1] < arr[i] && arr[i] >= arr[i+1]
unless edge_plateau?(arr, i)
peak_set[:pos] << i
peak_set[:peaks] << arr[i]
end
end
end
peak_set_alt = peak_set.collect{|k,v| [k.to_s, v]}.to_h
peak_set_alt
end
def edge_plateau?(array, position)
edge_plateau_left = true
edge_plateau_right = true
i = 1
until i == position
edge_plateau_left = false if array[0] != array[i]
i += 1
end
i = array.length-2
until i == position
edge_plateau_right = false if array[i] != array.last
i -= 1
end
edge_plateau_left or edge_plateau_right
end
Here's the test that it needs to pass but I don't know the original array, so that's a bit of a challenge.
Expected: {"pos"=>[2, 7, 14, 20], "peaks"=>[5, 6, 5, 5]}, instead got: {"pos"=>[2, 7, 11, 14, 20], "peaks"=>[5, 6, 3, 5, 5]}
I'm getting an extra peak in the middle but that should be ok if it's a local maxima, right?
UPDATE
Thanks to a suggestion I found the test array
[1, 2, 5, 4, 3, 2, 3, 6, 4, 1, 2, 3, 3, 4, 5, 3, 2, 1, 2, 3, 5, 5, 4, 3]

Answer 1

This is a more Ruby-like way to find the local maxima.
Code
def locale_maxima(arr)
last_idx = arr.size - 1
peaks, pos =
([[-Float::INFINITY, nil]] +
arr.each_with_index.reject { |v,i| i < last_idx && v == arr[i+1] } +
[[-Float::INFINITY, nil]]
).each_cons(3).
select { |(n1,_), (n2,_), (n3,_)| n1 < n2 && n2 > n3 }.
map { |_,max_pair,_| max_pair }.
transpose
{ pos: pos, peaks: peaks }
end
Example
arr = [1, 2, 5, 4, 3, 2, 3, 6, 4, 1, 2, 3, 3, 4, 5, 3, 2, 1, 2, 3, 5, 5, 4, 3]
locale_maxima arr
#=> { :pos =>[2, 7, 14, 21],
# :peaks=>[5, 6, 5, 5] }
Explanation
The steps are as follows.
last_idx = arr.size - 1
#=> 23
Where there are consecutive equal values, which may represent inflection points (a complication), remove all but the last. To report the indices of the local maxima we therefore need to save indices before removing the duplicates.
b = arr.each_with_index.reject { |v,i| i < last_idx && v == arr[i+1] }
#=> [[1, 0], [2, 1], [5, 2], [4, 3], [3, 4], [2, 5], [3, 6], [6, 7],
# [4, 8], [1, 9], [2, 10], [3, 12], [4, 13], [5, 14], [3, 15],
# [2, 16], [1, 17], [2, 18], [3, 19], [5, 21], [4, 22], [3, 23]]
Notice that [3,11] and [5, 20] have been removed.
Tack on pairs at the beginning and end that cannot be local maxima (nil is arbitrary).
c = [[-Float::INFINITY, nil]] + b + [[-Float::INFINITY, nil]]
#=> [[-Infinity, nil], [1, 0], [2, 1], [5, 2], [4, 3], [3, 4], [2, 5], [3, 6],
# [6, 7], [4, 8], [1, 9], [2, 10], [3, 12], [4, 13], [5, 14], [3, 15], [2, 16],
# [1, 17], [2, 18], [3, 19], [5, 21], [4, 22], [3, 23], [-Infinity, nil]]
Use Enumerable#each_cons to produce an enumerator that will be used to identify the local maxima.
d = c.each_cons(3)
#=> #<Enumerator:
# [[-Infinity, nil], [1, 0], [2, 1], [5, 2], [4, 3], [3, 4], [2, 5],
# [3, 6], [6, 7], [4, 8], [1, 9], [2, 10], [3, 12], [4, 13], [5, 14],
# [3, 15], [2, 16], [1, 17], [2, 18], [3, 19], [5, 21], [4, 22],
# [3, 23], [-Infinity, nil]]:each_cons(3)>
e = d.select { |(n1,_), (n2,_), (n3,_)| n1 < n2 && n2 > n3 }
#=> [[[2, 1], [5, 2], [4, 3]],
# [[3, 6], [6, 7], [4, 8]],
# [[4, 13], [5, 14], [3, 15]],
# [[3, 19], [5, 21], [4, 22]]]
f = e.map { |_,max_pair,_| max_pair }
#=> [[5, 2], [6, 7], [5, 14], [5, 21]]
peaks, pos = f.transpose
#=> [[5, 6, 5, 5], [2, 7, 14, 21]]
{ pos: pos, peaks: peaks }
#=> {:pos=>[2, 7, 14, 21], :peaks=>[5, 6, 5, 5]}

Answer 2

Well there are more than one bug in your code.
Try your code on following data
pick_peaks([0,1,10,1,2,2,3,1,10,1,0])
You'll get
{"pos"=>[2, 4, 6, 8], "peaks"=>[10, 2, 3, 10]}
Obviously 2 here is a bug. So the source of the bug here is arr[i] >= arr[i+1]
Also unless it is shomehow explicitly stated in the task, you seem to handle edges wrong. Consider
pick_peaks([0,0,10,1,2,2,3,1,10,0,0])
You'll get
{"pos"=>[4, 6], "peaks"=>[2, 3]}
missing both 10 on the left and on the right.
Without exact task it is hard to say definitely but at the first glance it seems that your algorithm is to complicated both in terms of code and Big-O. Why you don't just go once through the array and track whether your are climbing up, staying plain or going down?
Update
Here is a piece of code to illustrate my last suggestion. I'm not good at Ruby so my example code would be in JavaScript so you can run it in your browser console:
function pick_peaks(arr) {
var prevUp = true; //let the start be a peak
var peaks = [];
var lastUpInd = 0;
for(var i = 0; i < arr.length-1; i++) {
if (arr[i] < arr[i+1]) {
prevUp = true;
lastUpInd = i + 1;
}
else if (arr[i] > arr[i+1]) {
if(prevUp) {
for(var j = lastUpInd; j <= i; j++) {
peaks.push([j, arr[j]]);
}
}
prevUp = false;
}
}
// additionally handle the end to let it be a peak
if(prevUp)
{
for(var j = lastUpInd; j <= i; j++) {
peaks.push([j, arr[j]]);
}
}
return peaks;
}
If you don't want the ends to be possible peaks just init prevUp with false and remove last if with inner for after a comment