NaN results and internal error in optimizer

[SV-97]

I just ran DeCAFS with default arguments on a data sample and found that the output contains mostly NaNs. The input sample consists of 150 quite "normal" floats (no extremely small or large values, no NaN, no inf).

Reproduction:

library(DeCAFS)

y <- c(0.007487091243194955, 0.11791772771897412, 0.274858774774672, 0.32048047282650544, 0.3757456660237557, 0.566730137881182, 0.6468999628291104, 0.6825335929668119, 0.6684561104986123, 0.6742599957935859, 0.7123965269877737, 0.789119111739368, 0.7754592668321887, 0.9455229561156615, 0.8888850229080866, 0.9579045591010575, 0.9684453261535323, 0.9841466930118666, 1.01946205123237, 1.045175796792769, 0.981351126931916, 1.054297252524191, 0.942791676871981, 0.993005029590164, 1.0180014152473884, 0.9773316292505307, 0.9370515582015042, 0.9287021264216976, 0.9537520897393428, 0.9893110549878226, 0.9244673162832447, 0.9326054143717626, 0.8343611130380136, 0.8470182924547867, 0.8261301563301607, 0.7793719088575547, 0.3180211446782288, 0.31175824054646195, 0.3262753637162709, 0.2604205289329914, 0.01116637952088971, 0.4460261887369531, 0.4376237711891152, 0.40939557003502136, 0.38355278151646516, 0.3546338592731786, 0.13139142282684407, 0.1564255346584222, 0.14793892322342267, 0.12304102096857022, 0.07136721501014419, -0.011492751541630042, 0.04752713244714551, 0.07963591318517446, -0.022611757851454417, 0.0532148617806642, 0.053020104207421204, 0.058268629892130434, -0.04286326365559641, -0.019304572629339233, -0.011617545351878995, -0.04855446191674602, 0.09363161575240134, -0.020188874136647012, 0.01669865574916081, -0.011731061835822452, 0.08055727879776026, 0.06779383612941671, 0.05530644154140523, -0.08341616114978281, 0.031224197872941218, 0.06546575582720798, 0.08360276013086411, 0.015998939168694153, 0.14560395375002605, -0.002525927877456216, 0.04891629102073212, 0.25296946817818194, 0.2330548078002823, 0.3683625343556855, 0.48797658559001716, 0.4478107015692193, 0.5049142223839364, 0.47609272107494005, 0.5485907170173352, 0.6753893937196748, 0.722192989992027, 0.14311353260324516, 0.11112483988056437, 0.3628065796077881, 0.3229085631895193, 0.44015585731060797, 0.3629064849762885, 0.3630253898732863, 0.6138177453092869, 0.6979561855502128, 0.6614044981461945, 0.6769893756120073, 0.6904405919697462, 0.7483956350476936, 0.9744134604476808, 1.0026648371212223, 0.9886874859073326, 0.9355764088164243, 1.012842682176877, 0.6792236359711198, 0.6029048140844963, 0.6531411971377318, 0.25848459521735445, 0.9482914438926076, 0.9016693943091647, 0.9391852465583588, 0.9443951812436269, 0.9600693785169708, 0.992180570176749, 0.9266091472567773, 1.0577838405262177, 1.0307100143601928, 1.0382449383276637, 1.0525269463466438, 1.0002502423574253, 0.9872844193998384, 0.8009263038737743, 0.65125631418937, 0.5551140437855852, 0.5168431042571614, 0.2773549149869592, 0.2174255747559068, 0.016568715911897434, 0.9127055613108187, 0.8930462624084491, 0.9320777447604038, 0.9221333560289896, 0.8590934257726004, 0.7491357494384808, 0.7508279518141973, 0.6777549103073315, 0.6633019713402686, 0.3425453035598096, 0.459988930413731, 0.4074590298163433, 0.3444540185022145, 0.3302176201837615, 0.3019037768284888, 0.2002920314026595, 0.13132875882196646, 0.03320156243391202, 0.09465993597838829, -0.020184327783061555, 0.07033136127845799)

res <- DeCAFS(y)
res # contains a bunch of NaN

and corresponding output for this:

$changepoints
  [1]   1   2   3   4   5   6   7   8   9  10  11  12  13  14  15  16  17  18
 [19]  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36
 [37]  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54
 [55]  55  56  57  58  59  60  61  62  63  64  65  66  67  68  69  70  71  72
 [73]  73  74  75  76  77  78  79  80  81  82  83  84  85  86  87  88  89  90
 [91]  91  92  93  94  95  96  97  98  99 100 101 102 103 104 105 106 107 108
[109] 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126
[127] 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144
[145] 145 146 147 148 149

$signal
  [1] NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN
 [19] NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN
 [37] NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN
 [55] NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN
 [73] NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN
 [91] NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN
[109] NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN
[127] NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN
[145] NaN NaN NaN NaN NaN NaN

$costFunction
  tau    l   u   a   b   c
1   1 -Inf Inf NaN NaN NaN

$modelParameters
$modelParameters$sdEta
[1] NaN

$modelParameters$sdNu
[1] 0.1509451

$modelParameters$phi
         
0.850643 


$data
  [1]  0.007487091  0.117917728  0.274858775  0.320480473  0.375745666
  [6]  0.566730138  0.646899963  0.682533593  0.668456110  0.674259996
 [11]  0.712396527  0.789119112  0.775459267  0.945522956  0.888885023
 [16]  0.957904559  0.968445326  0.984146693  1.019462051  1.045175797
 [21]  0.981351127  1.054297253  0.942791677  0.993005030  1.018001415
 [26]  0.977331629  0.937051558  0.928702126  0.953752090  0.989311055
 [31]  0.924467316  0.932605414  0.834361113  0.847018292  0.826130156
 [36]  0.779371909  0.318021145  0.311758241  0.326275364  0.260420529
 [41]  0.011166380  0.446026189  0.437623771  0.409395570  0.383552782
 [46]  0.354633859  0.131391423  0.156425535  0.147938923  0.123041021
 [51]  0.071367215 -0.011492752  0.047527132  0.079635913 -0.022611758
 [56]  0.053214862  0.053020104  0.058268630 -0.042863264 -0.019304573
 [61] -0.011617545 -0.048554462  0.093631616 -0.020188874  0.016698656
 [66] -0.011731062  0.080557279  0.067793836  0.055306442 -0.083416161
 [71]  0.031224198  0.065465756  0.083602760  0.015998939  0.145603954
 [76] -0.002525928  0.048916291  0.252969468  0.233054808  0.368362534
 [81]  0.487976586  0.447810702  0.504914222  0.476092721  0.548590717
 [86]  0.675389394  0.722192990  0.143113533  0.111124840  0.362806580
 [91]  0.322908563  0.440155857  0.362906485  0.363025390  0.613817745
 [96]  0.697956186  0.661404498  0.676989376  0.690440592  0.748395635
[101]  0.974413460  1.002664837  0.988687486  0.935576409  1.012842682
[106]  0.679223636  0.602904814  0.653141197  0.258484595  0.948291444
[111]  0.901669394  0.939185247  0.944395181  0.960069379  0.992180570
[116]  0.926609147  1.057783841  1.030710014  1.038244938  1.052526946
[121]  1.000250242  0.987284419  0.800926304  0.651256314  0.555114044
[126]  0.516843104  0.277354915  0.217425575  0.016568716  0.912705561
[131]  0.893046262  0.932077745  0.922133356  0.859093426  0.749135749
[136]  0.750827952  0.677754910  0.663301971  0.342545304  0.459988930
[141]  0.407459030  0.344454019  0.330217620  0.301903777  0.200292031
[146]  0.131328759  0.033201562  0.094659936 -0.020184328  0.070331361

attr(,"class")
[1] "DeCAFSout" "list"    

While trying to reduce this data to a minimal reproducing example I further encountered an internal error:

Error in optim(par = start, .MoMCost, lower = c(phiLower, 0.001, 0), upper = c(phiUpper,  : 
  L-BFGS-B needs finite values of 'fn'

The reproducing example for this is:

library(DeCAFS)

y <- c(0.007487091243194955, 0.11791772771897412, 0.274858774774672, 0.32048047282650544, 0.3757456660237557, 0.566730137881182, 0.6468999628291104, 0.6825335929668119, 0.6684561104986123, 0.6742599957935859, 0.7123965269877737, 0.789119111739368, 0.7754592668321887, 0.9455229561156615, 0.8888850229080866)

res <- DeCAFS(y)

I originally ran into this issue on a Fedora Silverblue system (Fedora 40, Kernel 6.10.12-200, running on x64 AMD 5900X) however a colleague was able to reproduce it on their non-linux machine as well.