randomGuy Posted October 20, 2013 Posted October 20, 2013 Now tell him what fpir means, like what does fpir of .06% on 1.2 billion mean? He is saying all along that this means that every individual's biometrics will match .06% of 1.2 billion =7.2 lac records on average. Tell him he is wrong.
The Outsider Posted October 20, 2013 Posted October 20, 2013 He is saying all along that this means that every individual's biometrics will match .06% of 1.2 billion =7.2 lac records on average. Tell him he is wrong. A 0.06% FPR on 1.2 billion people means there are going to be 720,000 false positives out of the population. To achieve that one needs an individual FPR of around 5*10^(-13) or a 40k population FPR of 2*10^(-8). The current number seems to be around 2*10^(-3). We are 5 orders of magnitude away from that level of accuracy. If the individual FPR was 0.06%, then what DM is saying would hold - the FPR of 1.2 billion people would be 7,200,000. In that post, I am trying to reconcile .0025% on 40k with .057% on 100 millions somehow . So trying to say how fpir on 80k(which we are able to derive intuitively) may be say0.0026% and not .005%. But how can that magically be the case? In post #97, you are denying the presented FPR definition with which you seem to be agreeing now, by citing sources which claim an FPR of 0.057% on 100 million people. Anyhow, the crux here is that the scheme as it stands on available data appears to be loaded with issues.
Crookbond Posted October 20, 2013 Posted October 20, 2013 He is saying all along that this means that every individual's biometrics will match .06% of 1.2 billion =7.2 lac records on average. Tell him he is wrong. That's not what I said!
randomGuy Posted October 20, 2013 Posted October 20, 2013 A 0.06% FPR on 1.2 billion people means there are going to be 720' date='000 false positives out of the population. [/quote'] Thanks for this. I have been saying this all along. I understand .0025% on 40k and 0.057% on 100 million (which are both UIDAI records), seem to be not consistent with each other which is why I said at the start tht 0.0025% on 40k looks bogus. (and later tried to somehow reconcile it with .057% on 100 million). Both of us have a difference there, fine. Now see DMs definition of fpir which i have been trying to correct from the beginning and thats my aim first -
randomGuy Posted October 20, 2013 Posted October 20, 2013 Check this official report from jan-2012 3. False Positive Identification Rate (FPIR): 0.057%. In practical terms, it means that at a run rate of 10 lakh enrolments a day, only about 570 cases need to be manually reviewed daily to ensure that no resident is erroneously denied an Aadhaar number. Although this number is expected to grow as the database size increases, it is not expected to exceed manageable values even at full enrolment of 120 crores. The UIDAI currently have a manual adjudication team that reviews and resolves such cases http://uidai.gov.in/images/FrontPageUpdates/role_of_biometric_technology_in_aadhaar_jan21_2012.pdf
Crookbond Posted October 20, 2013 Posted October 20, 2013 Check this official report from jan-2012 http://uidai.gov.in/images/FrontPageUpdates/role_of_biometric_technology_in_aadhaar_jan21_2012.pdf First, this backs up my interpretation/calculations of FPIR in the earlier post. Page 17 from the report When new enrollment data from a resident is sent to CIDR, the system de-duplicates the resident packet to ensure that the resident has not previously been given a Aadhaar number. If the biometric de-duplication system rejects the new enrollment as being a duplicate, it is checked by manual adjudication process. If biometric system makes lots of errors, the volume of cases to adjudicate can go up significantly. This error by the biometric system is called false rejection of Aadhaar or False Positive Identification Rate (FPIR) of the biometric system. This is unlike what you said earlier - It simply means that - 0.06% of 1.2billion = 7,20,000 people will NOT be able to enroll because their biometric matches falsely with some other unique record. So, basically 7.2 lac of 12,000 lac (i.e. 1.2 billion) people will not be able to enroll. This figure would be acceptable to most concerned with UID as well as with the gov. The metric number you are looking for is FER or Failure to Enroll Rate. However, this number is set to 0 - you can't have people not getting an Aadhar since it is a "right". Page 17 of the report - Per UIDAI policy, failure to enroll in Aadhaar is not allowed. That is, Aadhaar is a right of every Indian resident and cannot be denied. Therefore overall failure to enroll is set by policy to be zero. However, if certain residents are not able to provide their biometric (called B-FTE henceforth), biometric de-duplication cannot be carried out on their enrollment. Therefore the UIDAI measures B-FTE on an ongoing basis However, unlike the report I told "How many comparisons" one needs to make to ensure no duplicates viz. resolve these 570 in 100,000 cases. The report only says 570 in 100,000 cases but doesn't say "how much manual effort" for each case. My earlier post gives some idea on the "amount of manual effort required" to handle those 570 cases i.e. for each of these 570 cases how many other records should I look at before ensuring that this is not a duplicate recrrd? The calculations show that there is a huge number of manual effort required. Mind you, deduplication can NOT be performed at Aadhar enrolment because the system is not real time. Second, This is equivalent to saying "No. 1 X in the world" (X= water cooler, washing machine, airport etc.). Of course, I will big up my system. you can't have an organization building the system evaluate it. The evaluation needs to be done by an independent authority on the subject. The veracity and the technical competence of this report is bound to open interpretation and not to be taken at face value. In fact, even if these numbers were taken at face value the numbers are so shoddy that the system will cease to exist.
Yoda-esque Posted October 20, 2013 Posted October 20, 2013 Why would anybody want to big up their system?
Crookbond Posted October 20, 2013 Posted October 20, 2013 Why would anybody want to big up their system? Because you built-it, it's your baby. You want to show it's perfect, it's best. No brainer really. Google feel they are the best search engine, Bing however says that they are. In any case, that is besides the point. Whether you want to big it up or not, the evaluation of your system has to be done by an independent authority on the subject and not by the person who built the system. The person who built the system would by definition (and common sense) have a bias towards it.
randomGuy Posted October 20, 2013 Posted October 20, 2013 DM, please read post 113 and please show some honestly to admit my definition is right (regarding false positives) and yours is wrong.
randomGuy Posted October 20, 2013 Posted October 20, 2013 Why would anybody want to big up their system? The system is actually good. Nearly 600 million have already come to enrol, nearly 460 million have already been alloted aadhar numbers. False positive rate is .057% for 8.5 crores its pretty acceptable, see post 113. There is simply no major criticism, DM is clutching at straws.
Crookbond Posted October 20, 2013 Posted October 20, 2013 DM' date=' please read post 113 and please show some honestly to admit my definition is right (regarding false positives) and yours is wrong.[/quote'] I have the read the entire report yesterday. In addition, I discussed this with a professor (Ph.D in Biometrics from one of the leading US university in Biometrics) plus two additional graduate students who are close to their completion of PhD in biometrics. They concur with my analysis. I repeat and here is what you said earlier about FPIR- I do not see any mention of this in the report. In fact, the report posted in #113 says Focus on the highlighted - You have messed up FER with FPIR. Btw, if you still want to hold with your definition of FPIR; I can settle this pretty easily. Let me know if you still want to take this further.
The Outsider Posted October 20, 2013 Posted October 20, 2013 Check this official report from jan-2012 http://uidai.gov.in/images/FrontPageUpdates/role_of_biometric_technology_in_aadhaar_jan21_2012.pdf A 0.057% FPR for 10 lakh people translates to a >60% FPR for 1.2 billion people. That's a crap system by any stretch of imagination.
randomGuy Posted October 20, 2013 Posted October 20, 2013 A 0.057% FPR for 10 lakh people translates to a >60% FPR for 1.2 billion people. That's a crap system by any stretch of imagination. That's for 8.5 crore, not for 10 lac. The report is saying how it works in practical terms on their 10lac registrations per day. 3. False Positive Identification Rate (FPIR): 0.057%. In practical terms, it means that at a run rate of 10 lakh enrolments a day, only about 570 cases need to be manually reviewed daily to ensure that no resident is erroneously denied an Aadhaar number. Although this number is expected to grow as the database size increases, it is not expected to exceed manageable values even at full enrolment of 120 crores. The UIDAI currently have a manual adjudication team that reviews and resolves such cases http://uidai.gov.in/images/FrontPageUpdates/role_of_biometric_technology_in_aadhaar_jan21_2012.pdf Read it. The system is actually good.
Crookbond Posted March 24, 2014 Posted March 24, 2014 Cobrapost exposes UID Scam http://cobrapost.com/index.php/news-detail?nid=5773&cid=23
Precambrian Posted March 24, 2014 Posted March 24, 2014 http://zeenews.india.com/news/nation/sc-asks-centre-to-withdraw-all-orders-making-aadhar-mandatory-for-govt-service_919886.html The Supreme Court on Monday directed the central government to withdraw all its orders making Aadhar mandatory for people for availing any government services. The three-judge bench headed by Justice BS Chauhan ordered the Unique Identification Authority of India (UIDAI) not to share any information of the Aadhar card holder with any government agency. So that's it. UPA can rest easy that they have managed to sell more than 60% of the population's core data to US agencies. :hitler:
Crookbond Posted September 8, 2014 Posted September 8, 2014 Aadhar shocker: Names suffixed with choicest abuse http://timesofindia.indiatimes.com/city/kolkata/Aadhar-shocker-Names-suffixed-with-choicest-abuse/articleshow/41971532.cms "After the surname Bhowmick, the word Bokac**** has been mentioned, and instead of my father's name, you will see the word 'Dog'," he told TOI. ... What's more, his gender has been mentioned as "transgender", just as it has been with Chittaranjan Sarkar, another resident of Baidyabati, though his father's name has been spared such an abomination
Recommended Posts