Video
Create Note
Dear user,
with the manuscript function you'll be able to create your own digital lecture manuscript.
However, in order to link all your notes with your user profile it is required that you
login to the tele-TASK portal to use this functionality.
If you don't have an account yet, you may register for a tele-TASK account here.
with the manuscript function you'll be able to create your own digital lecture manuscript.
However, in order to link all your notes with your user profile it is required that you
login to the tele-TASK portal to use this functionality.
If you don't have an account yet, you may register for a tele-TASK account here.
Add a New Marker
Dear user,
with the marker function you'll be able to create your own digital time markers.
However, in order to link all your markers with your user profile it is required that you
login to the tele-TASK portal to use this functionality.
If you don't have an account yet, you may register for a tele-TASK account here.
with the marker function you'll be able to create your own digital time markers.
However, in order to link all your markers with your user profile it is required that you
login to the tele-TASK portal to use this functionality.
If you don't have an account yet, you may register for a tele-TASK account here.
Tag this content
Please enable javascript to use this function.
Dear user,with the tagging function you'll be able to add taggs to videos.
However, in order to link all your tags with your user profile it is required that you
login to the tele-TASK portal to use this functionality.
If you don't have an account yet, you may register for a tele-TASK account here.
Tags added to this content
No tags have been added to this content so far.
Add Link to this content
Please enable javascript to use this function.
Dear user,with the links function you'll be able to add links to other resources to this content.
However, in order to link all your links with your user profile it is required that you
login to the tele-TASK portal to use this functionality.
If you don't have an account yet, you may register for a tele-TASK account here.
Links added to this content
No links have been added to this content so far.
Add to my playlist
Dear user,
with the playlist function you'll be able to create your own lecture video playlists.
However, in order to link all your playlists with your user profile it is required that you
login to the tele-TASK portal to use this functionality.
If you don't have an account yet, you may register for a tele-TASK account here.
with the playlist function you'll be able to create your own lecture video playlists.
However, in order to link all your playlists with your user profile it is required that you
login to the tele-TASK portal to use this functionality.
If you don't have an account yet, you may register for a tele-TASK account here.
Playlists
This content is not used in any playlist.
144971873674215486206268771031103310681167120314221491149518821883197622052211225224302725280229212998299932453386350736213772380141144136413741984199426344224447449044934494449545674568456945984759493250665068
Lecture Structure
- Contents (00:00:44)
- Introduction (00:00:44)
- happen (00:01:37)
- Wise men are aware of future things just about to (00:01:37)
- lone are totally enlightened (00:01:37)
- gods know what the future holds because they (00:01:37)
- Ordinary mortals know what s h (00:01:37)
- Motivation (00:03:07)
- Ö rpa cti ve Fault Management ( (00:07:01)
- ob ecuves (00:07:01)
- Classical reliability theory may help but is not very good for (00:09:08)
- Variable Selection (00:10:20)
- Combir atorial explosionl (00:10:26)
- For n nodes m = v X fx n variables the number of (00:10:26)
- Variable Selection Methods (00:14:37)
- mäß (00:17:11)
- on time series and (00:17:11)
- PWA performs best (00:17:11)
- Variables (00:17:11)
- techniques (00:17:11)
- Ber chmarked four (00:17:11)
- Variable Selection (00:17:48)
- class label data (00:17:48)
- on time series and (00:17:48)
- PWA performs best (00:17:48)
- Variables (00:17:48)
- techniques (00:17:48)
- Ber chmarked four (00:17:48)
- Contents (00:19:27)
- Fau ts Errors Failures a (00:20:03)
- Fau ts Errors Failures again (00:23:42)
- Fou Ways of Detecting Fau ts (00:24:51)
- Vaxonom y (00:24:55)
- Reporting (00:24:55)
- Tracking (00:24:55)
- Sy fgptqm (00:24:55)
- Undetected (00:24:55)
- Detected (00:24:55)
- the current situation is (00:31:22)
- a decision that failure is n minent or not or (00:31:22)
- Sy qptqm (00:31:22)
- On ine Failure Prediction De nition (00:31:23)
- based on runtime monitoring data (00:31:23)
- fail fre prone situation i e situations that will (00:31:23)
- The goal of online failure prediction is to identify (00:31:23)
- data window (00:32:56)
- data (00:32:56)
- Examples for event based (00:32:56)
- Examples for periodic data (00:32:56)
- event based C GQOF CE (00:32:56)
- periodic numerical (00:32:56)
- There are two types of (00:32:56)
- Contents (00:36:45)
- Prediction Techniques Examples (00:36:51)
- UBF Background (00:45:24)
- Dispersion Frame fechnique (00:46:42)
- mäß (00:48:41)
- containing a database of indicative eventsets (00:48:41)
- Result rule based failure prediction system (00:48:41)
- Based on sets of events (00:48:41)
- Approach inspired by data mining (00:48:41)
- nal evenfs ldatabase (00:49:58)
- nal eventset (00:49:59)
- Initial eventset (00:49:59)
- failure window 2 (00:49:59)
- failure window 1 (00:49:59)
- Hidden Semi Markov Model Prediction (00:54:05)
- Fault tolerant systems (00:54:05)
- nds onOdepe we (00:54:05)
- Dependencies lead to error patterns (00:54:05)
- pproach (00:56:26)
- Development of a Hidden Semi Markov Model (HSMM) (00:56:26)
- Additional assumption (00:56:26)
- Machine learning (00:56:26)
- Hidden Semi Markov Models (00:58:27)
- Machine Learning wo Steps (01:00:21)
- ß ecision r Recall and other Metrics (01:03:21)
- True positive rate is equal to recall (01:03:21)
- False positive rate (fpr) (01:03:21)
- Recall fraction of predicted failures (01:03:21)
- Precision fract on of correct alarms P1 (01:03:21)
- No warning (01:03:21)
- Recei ver Operating v Characteristics (01:08:34)
- Threshold oo tpr and fpr equal to one (01:08:34)
- Threshold oo tpr and fpr equal to zero (01:08:34)
- Plot true positive rate (recall) over false positive rate for (01:08:34)
- Q ecision (01:08:56)
- True positive rate s equal to recall (01:08:56)
- False positive rate ( (01:08:56)
- Recall fraction of predicted failures (01:08:56)
- Precision fractior of correct alarrns P1 (01:08:56)
- Machine Learning wo Steps (01:08:57)
- Contents (01:09:58)
- lf gecision r Recall and other Metrics (01:09:59)
- True positive rate is equal to recall (01:09:59)
- Recall fraction of predicted failures (01:09:59)
- Precision fraction of correct alarms P1 (01:09:59)
- No warning (01:09:59)
- jReceiver Operating v Characteristics (01:11:04)
- Threshold oo tpr and fpr equal to one (01:11:04)
- Threshold oo tpr and fpr equal to zero (01:11:04)
- Plot true positive rate (recall) over false positive rate for (01:11:04)
- Precision Recal Plots (01:13:42)
- Precision Recall Plots (01:14:07)
- predictor (01:14:07)
- Goal obtain one real number to evaluate quality of (01:14:07)
- Good for visual inspection bad for algorithmic decisions (01:14:07)
- Contents (01:14:50)
- mäß (01:14:53)
- (Auc) (01:14:53)
- predictor (01:14:53)
- Goal obtain one real number to evaluate quality of (01:14:53)
- Good for visual inspection bad for algorithmic decisions (01:14:53)
- Precision Reca P ots (01:14:54)
- neqative examples recall equal to one (01:14:54)
- Threshold oo precision equal to ratio of positive and (01:14:54)
- Threshold oo precision equal to one recall equal to zero (01:14:54)
- Plot pre cision over recall for various thresholds (01:14:54)
- Receiver Operating v Characteristics (ROC) (01:14:55)
- Threshold oo tpr and fpr equal to one (01:14:55)
- Threshold oo tpr and fpr equal to zero (01:14:55)
- Plot true positive rate (recall) over false positive rate for (01:14:55)
- Precision Reca P ots (01:16:07)
- Threshold oo precision equal to one recall equal to zero (01:16:07)
- Plot precision over recall for various thresholds (01:16:07)
- Ska ar Metrics (01:16:08)
- predictor (01:16:08)
- Goal obtain one real number to evaluate quality of (01:16:08)
- Good for visual inspection bad for algorithmic decisions (01:16:08)
- Contents (01:16:09)
- Taking Action (01:16:38)
- In general the following steps have to be performed (01:16:38)
- be taken in order to (01:16:38)
- After a potential failure has been predicted an action must (01:16:38)
- proactively (01:16:38)
- Taxonomy of Reaction Methods (01:19:19)
- Effects of Proactive Methods (01:22:12)
- vailabilityl (01:22:12)
- However In case of frequent false positive and false negative (01:22:12)
- Downtime minimization reduces M R (01:22:12)
- Introduction (01:24:26)
- Summary (01:27:45)
Keyword
Please enable javascript to use this function.
Please enable javascript to use this function.