zfzf21 / BASIC_changepoint

Software implementation of the BASIC changepoint model and inference algorithms
0 stars 1 forks source link

Fatal errors #1

Closed ivukotic closed 7 years ago

ivukotic commented 7 years ago

Hi,

I have tried using your code but run into a bit of problem... here's output: `MCEM Step...

** FATAL ERROR DETECTED ** ** FATAL ERROR DETECTED ** ** FATAL ERROR DETECTED **

Two fatal errors have been detected, the first was inappropriately ignored. To prevent further fatal errors from being ignored this application will be terminated immediately and you should go fix this buggy program.

The error message from this fatal error was:

Error detected at line 3386. Error detected in file ../src/dlib/optimization/optimization_bobyqa.h. Error detected in function double dlib::find_min_bobyqa(const funct&, T&, long int, const U&, const U&, double, double, long int) [with funct = dlib::negate_function_object<model_params_objective >; T = dlib::matrix<double, 0, 1>; U = dlib::matrix<double, 0, 1>].

Failing expression was x.size() + 2 <= npt && npt <= (x.size()+1)(x.size()+2)/2 && 0 < rho_end && rho_end < rho_begin && min(x_upper - x_lower) > 2rho_begin && min(x - x_lower) >= 0 && min(x_upper - x) >= 0. void find_min_bobyqa() Invalid arguments have been given to this function ntp in valid range: true npt: 7 rho_begin: 10 rho_end: 1e-06 min(x_upper - x_lower) > 2*rho_begin: true min(x - x_lower) >= 0 && min(x_upper - x) >= 0: false`

It could be that there is something strange in my data, please let me know if you would like me to send you my data so you could reproduce the problem.

Thanks, Ilija

zfzf21 commented 7 years ago

Hi Ilija,

Thanks for the notification about this error. Yes, can you send me your data, and also how you are running the program (i.e. the specific Python call you used, if you are using the Python interface)? I'll try to reproduce the problem on my machine.

I (and several others) have gotten this program to install and run successfully on several different data sets, but it hasn't been tested across a broad set of inputs and applications so it's possible you found a robustness problem.

Best, Zhou

On Fri, May 26, 2017 at 9:00 AM, Ilija Vukotic notifications@github.com wrote:

Hi,

I have tried using your code but run into a bit of problem... here's output: `MCEM Step...

** FATAL ERROR DETECTED ** ** FATAL ERROR DETECTED ** ** FATAL ERROR DETECTED **

Two fatal errors have been detected, the first was inappropriately ignored. To prevent further fatal errors from being ignored this application will be terminated immediately and you should go fix this buggy program.

The error message from this fatal error was:

Error detected at line 3386. Error detected in file ../src/dlib/optimization/optimization_bobyqa.h. Error detected in function double dlib::find_min_bobyqa(const funct&, T&, long int, const U&, const U&, double, double, long int) [with funct = dlib::negate_function_object<model_params_objective

; T = dlib::matrix<double, 0, 1>; U = dlib::matrix<double, 0, 1>].

Failing expression was x.size() + 2 <= npt && npt <= (x.size()+1)(x.size()+2)/2 && 0 < rho_end && rho_end < rho_begin && min(x_upper - x_lower) > 2rho_begin && min(x - x_lower) >= 0 && min(x_upper - x) >= 0. void find_min_bobyqa() Invalid arguments have been given to this function ntp in valid range: true npt: 7 rho_begin: 10 rho_end: 1e-06 min(x_upper - x_lower) > 2*rho_begin: true min(x - x_lower) >= 0 && min(x_upper - x) >= 0: false`

It could be that there is something strange in my data, please let me know if you would like me to send you my data so you could reproduce the problem.

Thanks, Ilija

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/zfzf21/BASIC_changepoint/issues/1, or mute the thread https://github.com/notifications/unsubscribe-auth/AGFzX3nO99KExGRhscUnimY6JQwn__Bqks5r9vc1gaJpZM4Nnxqf .

ivukotic commented 7 years ago

Hi Zhou,

here the data and a very simple test case: https://www.amazon.com/clouddrive/share/fZpdl3dhKRsmVpJXcUNuztrNY747JU0gdTTp1QU2HQx?ref_=cd_ph_share_link_copy Please let me know if it is really a bug or I just did something stupid.

Thanks, Ilija

-- Nullius in verba

On May 26, 2017, at 15:45 , zfzf21 notifications@github.com<mailto:notifications@github.com> wrote:

Hi Ilija,

Thanks for the notification about this error. Yes, can you send me your data, and also how you are running the program (i.e. the specific Python call you used, if you are using the Python interface)? I'll try to reproduce the problem on my machine.

I (and several others) have gotten this program to install and run successfully on several different data sets, but it hasn't been tested across a broad set of inputs and applications so it's possible you found a robustness problem.

Best, Zhou

On Fri, May 26, 2017 at 9:00 AM, Ilija Vukotic notifications@github.com<mailto:notifications@github.com> wrote:

Hi,

I have tried using your code but run into a bit of problem... here's output: `MCEM Step...

** FATAL ERROR DETECTED ** ** FATAL ERROR DETECTED ** ** FATAL ERROR DETECTED **

Two fatal errors have been detected, the first was inappropriately ignored. To prevent further fatal errors from being ignored this application will be terminated immediately and you should go fix this buggy program.

The error message from this fatal error was:

Error detected at line 3386. Error detected in file ../src/dlib/optimization/optimization_bobyqa.h. Error detected in function double dlib::find_min_bobyqa(const funct&, T&, long int, const U&, const U&, double, double, long int) [with funct = dlib::negate_function_object<model_params_objective

; T = dlib::matrix<double, 0, 1>; U = dlib::matrix<double, 0, 1>].

Failing expression was x.size() + 2 <= npt && npt <= (x.size()+1)(x.size()+2)/2 && 0 < rho_end && rho_end < rho_begin && min(x_upper - x_lower) > 2rho_begin && min(x - x_lower) >= 0 && min(x_upper - x) >= 0. void find_min_bobyqa() Invalid arguments have been given to this function ntp in valid range: true npt: 7 rho_begin: 10 rho_end: 1e-06 min(x_upper - x_lower) > 2*rho_begin: true min(x - x_lower) >= 0 && min(x_upper - x) >= 0: false`

It could be that there is something strange in my data, please let me know if you would like me to send you my data so you could reproduce the problem.

Thanks, Ilija

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/zfzf21/BASIC_changepoint/issues/1, or mute the thread https://github.com/notifications/unsubscribe-auth/AGFzX3nO99KExGRhscUnimY6JQwn__Bqks5r9vc1gaJpZM4Nnxqf .

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://github.com/zfzf21/BASIC_changepoint/issues/1#issuecomment-304384283, or mute the threadhttps://github.com/notifications/unsubscribe-auth/ABqBGHKioVm1L2t1WcQqrXuDgEYKPqcIks5r9znjgaJpZM4Nnxqf.

zfzf21 commented 7 years ago

Hi Ilija,

The error is occurring because your input data matrix X has missing values (represented by np.nan). Unfortunately the current code does not support missing values in the data. As a work-around, you will need to impute the missing values yourself before fitting the model.

I have pushed a minor update which checks for missing and infinite values, and stops execution immediately if such values are found. (So the program would still fail for your data, but at an earlier stage and with an explicit error message.)

Best, Zhou