{ "id": 8256, "title": "Problems with freq and latency in Octave 3.8.1", "url": "/competitions/seizure-detection/discussion/8256", "author_name": "", "post_date": "2014-05-22T19:09:01.663Z", "votes": 1, "comment_count": 10, "views": 2290, "content": "

I've tried loading some of the .mat files in Octave 3.8.1.  The data and channels variables look fine, but latency and freq seem to be empty matrices (1x0 matrices).  Any suggestions?

\n

Is anyone using Octave having success getting at these two fields?

", "messages": [ { "id": "46092", "postDate": "05/22/2014 19:09:01", "content": "

I've tried loading some of the .mat files in Octave 3.8.1.  The data and channels variables look fine, but latency and freq seem to be empty matrices (1x0 matrices).  Any suggestions?

\n

Is anyone using Octave having success getting at these two fields?

", "rawMarkdown": "", "votes": null }, { "id": "46121", "postDate": "05/23/2014 13:21:11", "content": "

Same here, nothing in those matrices.  My understanding is that it should be for ictal entries.

", "rawMarkdown": "", "votes": null }, { "id": "46137", "postDate": "05/23/2014 16:29:07", "content": "

Jeffrey is right - latency is only in ictal entries (the training files).

\n

If it helps, output from MATLAB is shown below for three files.

\n\n

Good luck!

\n

>> d1_s1_1 = load('Dog_1_ictal_segment_1');
>> d1_s1_2 = load('Dog_1_ictal_segment_2');

\n

>> d1_s1_1

\n

d1_s1_1 =data: [16x400 double]
freq: 399.6098
channels: [1x1 struct]
latency: 0

\n

>> d1_s1_2

\n

d1_s1_2 =data: [16x400 double]
freq: 399.6098
channels: [1x1 struct]
latency: 1

\n

>> d1_s0_1 = load('Dog_1_interictal_segment_1.mat');

\n

>> d1_s0_1

\n

d1_s0_1 =data: [16x400 double]
freq: 399.6098
channels: [1x1 struct]

", "rawMarkdown": "", "votes": null }, { "id": "47081", "postDate": "05/26/2014 12:23:03", "content": "

I'm with Eric on this one.  I did exactly what zeros showed in his post and there was nothing in freq or latency.  It shows:

\n

freq = [](1x0)

\n

and 

\n

latency = [](1x0)

\n

Channels and data has what it is supposed to.  Any ideas?

", "rawMarkdown": "", "votes": null }, { "id": "47092", "postDate": "05/26/2014 15:45:51", "content": "

Update: I just spent a few hours getting homebrew on my mac to install the latest ocatave 3.8.1 but it still doesn't have anything in any of the freq or latency values.  Anyone having luck with other tools?  Later today, I'll try reading the info with other tools and see if I get any luck.

\n

Would I be correct in assuming that if I were able to explode the data, it is probably not corrupt?

", "rawMarkdown": "", "votes": null }, { "id": "47100", "postDate": "05/26/2014 18:46:47", "content": "

I switched to using R to read the data and that's been working for me.  No problems with the latency or frequency fields.  There's another post in the forum that describes the package you need - I think it's "R.matlab" or something.

\n

Jeffrey, there's no chance your issues (our issues) with latency and frequency are due to data corruption.

", "rawMarkdown": "", "votes": null }, { "id": "47108", "postDate": "05/26/2014 23:37:18", "content": "

R did the trick! Eric, I really appreciate the help on this.  It is very frustrating getting over the hurdles of big data, but WOW is it a fascinating subject.  Hope I can help you out sometime.

", "rawMarkdown": "", "votes": null }, { "id": "49137", "postDate": "06/14/2014 10:51:11", "content": "

Hi,

\n

Same issue here.

\n

The issue might be with octave, but in any case it is useful to know with what version of matlab where the files created. Is it matlab v7 format?

\n

If you are using newer releases of matlab there are several datatypes that depend on classes (are latency and freq just doubles?). This of course makes compatibility quite difficult.

", "rawMarkdown": "", "votes": null }, { "id": "49168", "postDate": "06/15/2014 21:41:17", "content": "

Hi,

\n

It is indeed a bug in octave. The files were created using matlab format v5.

\n

Sadly, loading from v5 in Octave 3.8.1 seems to have a bug.

\n

If you still want to use Octave here is a workaround until the bug is fixed.

\n

Since scipy doesn't have this bug, if you are patient you can update all files to v7 format using the following python script

\n\n

from scipy.io import loadmat, savemat
from os import listdir
from os.path import isdir, isfile, join

\n

import sys

\n

mypath = "clips"

\n

folders = [ f for f in listdir(mypath) if isdir(join(mypath,f)) ]

\n

for folder in folders:
  tmp = join(mypath,folder)
  files = [ f for f in listdir(tmp) if isfile(join(tmp,f)) ]
 

\n

  for dataf in files:
    thepath = join (tmp,dataf)
    data = loadmat (thepath)
    print thepath
    savemat (thepath, data)
    sys.stdout.flush()

\n\n

Be warned: this will take a while. 

\n

After this Octave 3.8.1 reads all the content of the file correctly.

\n\n

UPDATE:

\n

If a file is saved in matlab with version 5 format; then  Octave can read it without a problem. 

\n

The issue is with v5 files created in R (which is the case for this data). 

\n

However other read-from-file functions seem to do the trick (like python).

\n

We are working to fix this...

", "rawMarkdown": "", "votes": null }, { "id": "49204", "postDate": "06/16/2014 15:21:25", "content": "

Wow, kakila, I have to give you major kudos!  Thanks for being a team player and figuring that out.

", "rawMarkdown": "", "votes": null }, { "id": "49208", "postDate": "06/16/2014 16:13:40", "content": "

The bug has been fixed. It will be part of the 3.8.2 bug fix release

\n\n

https://savannah.gnu.org/bugs/?42562

\n\n

It seems though it is a glitch in R, we will let them know

", "rawMarkdown": "", "votes": null } ], "comments": [ { "id": 46121, "author_name": "jeffrichley", "author_url": "", "post_date": "05/23/2014 13:21:11", "content": "

Same here, nothing in those matrices.  My understanding is that it should be for ictal entries.

", "votes": null, "replies": [] }, { "id": 46137, "author_name": "asaini", "author_url": "", "post_date": "05/23/2014 16:29:07", "content": "

Jeffrey is right - latency is only in ictal entries (the training files).

\n

If it helps, output from MATLAB is shown below for three files.

\n\n

Good luck!

\n

>> d1_s1_1 = load('Dog_1_ictal_segment_1');
>> d1_s1_2 = load('Dog_1_ictal_segment_2');

\n

>> d1_s1_1

\n

d1_s1_1 =data: [16x400 double]
freq: 399.6098
channels: [1x1 struct]
latency: 0

\n

>> d1_s1_2

\n

d1_s1_2 =data: [16x400 double]
freq: 399.6098
channels: [1x1 struct]
latency: 1

\n

>> d1_s0_1 = load('Dog_1_interictal_segment_1.mat');

\n

>> d1_s0_1

\n

d1_s0_1 =data: [16x400 double]
freq: 399.6098
channels: [1x1 struct]

", "votes": null, "replies": [] }, { "id": 47081, "author_name": "jeffrichley", "author_url": "", "post_date": "05/26/2014 12:23:03", "content": "

I'm with Eric on this one.  I did exactly what zeros showed in his post and there was nothing in freq or latency.  It shows:

\n

freq = [](1x0)

\n

and 

\n

latency = [](1x0)

\n

Channels and data has what it is supposed to.  Any ideas?

", "votes": null, "replies": [] }, { "id": 47092, "author_name": "jeffrichley", "author_url": "", "post_date": "05/26/2014 15:45:51", "content": "

Update: I just spent a few hours getting homebrew on my mac to install the latest ocatave 3.8.1 but it still doesn't have anything in any of the freq or latency values.  Anyone having luck with other tools?  Later today, I'll try reading the info with other tools and see if I get any luck.

\n

Would I be correct in assuming that if I were able to explode the data, it is probably not corrupt?

", "votes": null, "replies": [] }, { "id": 47100, "author_name": "onemillionmonkeys", "author_url": "", "post_date": "05/26/2014 18:46:47", "content": "

I switched to using R to read the data and that's been working for me.  No problems with the latency or frequency fields.  There's another post in the forum that describes the package you need - I think it's "R.matlab" or something.

\n

Jeffrey, there's no chance your issues (our issues) with latency and frequency are due to data corruption.

", "votes": null, "replies": [] }, { "id": 47108, "author_name": "jeffrichley", "author_url": "", "post_date": "05/26/2014 23:37:18", "content": "

R did the trick! Eric, I really appreciate the help on this.  It is very frustrating getting over the hurdles of big data, but WOW is it a fascinating subject.  Hope I can help you out sometime.

", "votes": null, "replies": [] }, { "id": 49137, "author_name": "kakila", "author_url": "", "post_date": "06/14/2014 10:51:11", "content": "

Hi,

\n

Same issue here.

\n

The issue might be with octave, but in any case it is useful to know with what version of matlab where the files created. Is it matlab v7 format?

\n

If you are using newer releases of matlab there are several datatypes that depend on classes (are latency and freq just doubles?). This of course makes compatibility quite difficult.

", "votes": null, "replies": [] }, { "id": 49168, "author_name": "kakila", "author_url": "", "post_date": "06/15/2014 21:41:17", "content": "

Hi,

\n

It is indeed a bug in octave. The files were created using matlab format v5.

\n

Sadly, loading from v5 in Octave 3.8.1 seems to have a bug.

\n

If you still want to use Octave here is a workaround until the bug is fixed.

\n

Since scipy doesn't have this bug, if you are patient you can update all files to v7 format using the following python script

\n\n

from scipy.io import loadmat, savemat
from os import listdir
from os.path import isdir, isfile, join

\n

import sys

\n

mypath = "clips"

\n

folders = [ f for f in listdir(mypath) if isdir(join(mypath,f)) ]

\n

for folder in folders:
  tmp = join(mypath,folder)
  files = [ f for f in listdir(tmp) if isfile(join(tmp,f)) ]
 

\n

  for dataf in files:
    thepath = join (tmp,dataf)
    data = loadmat (thepath)
    print thepath
    savemat (thepath, data)
    sys.stdout.flush()

\n\n

Be warned: this will take a while. 

\n

After this Octave 3.8.1 reads all the content of the file correctly.

\n\n

UPDATE:

\n

If a file is saved in matlab with version 5 format; then  Octave can read it without a problem. 

\n

The issue is with v5 files created in R (which is the case for this data). 

\n

However other read-from-file functions seem to do the trick (like python).

\n

We are working to fix this...

", "votes": null, "replies": [] }, { "id": 49204, "author_name": "jeffrichley", "author_url": "", "post_date": "06/16/2014 15:21:25", "content": "

Wow, kakila, I have to give you major kudos!  Thanks for being a team player and figuring that out.

", "votes": null, "replies": [] }, { "id": 49208, "author_name": "kakila", "author_url": "", "post_date": "06/16/2014 16:13:40", "content": "

The bug has been fixed. It will be part of the 3.8.2 bug fix release

\n\n

https://savannah.gnu.org/bugs/?42562

\n\n

It seems though it is a glitch in R, we will let them know

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "46092": "", "46121": "", "46137": "", "47081": "", "47092": "", "47100": "", "47108": "", "49137": "", "49168": "", "49204": "", "49208": "" }, "source": "meta" }