The authors introduce a new version of the MNIST data set that they call QMNIST, which is the result of a very thoughtful and systematic analysis of existing materials used to build the original MNIST. I am impressed by the meticulous investigation carried out to recover the precise processing steps needed to generate MNIST examples from the original NIST images. The QMNIST data set is thus the product of a very accurate reconstruction of the original process (though the authors note that some minor discrepancies are still present). The authors then investigate whether the performance of popular classification methods measured using the new QMNIST test set actually differ from that measured on the original MNIST test set. Overall, I think this research is well conducted and presented in a very clear way.

Although the popular MNIST dataset \citep{mnist} is derived from the NIST database \citep{nist-sd19}, precise processing steps of this derivation have been lost to time. We propose a reconstruction that is accurate enough to serve as a replacement for the MNIST dataset, with insignificant changes in accuracy. We trace each MNIST digit to its NIST source and its rich metadata such as writer identifier, partition identifier, etc. We also reconstruct the complete MNIST test set with 60,000 samples instead of the usual 10,000. Since the balance 50,000 were never distributed, they enable us to investigate the impact of twenty-five years of MNIST experiments on the reported testing performances.

Thousands of cold cases could be solved thanks to an breakthrough in fingerprint analysis by artificial intelligence. A computer using artificial intelligence system has shattered the received wisdom of decades that each fingerprint from a person's finger is unique. So if a criminal left a thumbprint at one crime scene, and a print from his index finger at another, there would be no way to link the two. The breakthrough came about when a Columbia University student attempted to see if artificial intelligence could find links between apparently very different fingerprints from the same person. To test the idea, Gabe Guo, an engineering graduate with no background in forensics presented a computer with images of some 60,000 fingerprints in pairs.

Harvey Castro talks about how AI could be used in cold cases and the symbiotic relationship between AI and a detective. New Jersey police are deploying new technology to try to break an unsolved case in what some experts believe could be the greatest advancement in cold-case investigations since forensic genetic genealogy caught the infamous Golden State Killer in 2018. A police department in the 70-square-mile town of Middle Township, along with the Cape May County Prosecutor's Office, will use artificial intelligence to try to solve the case of Mark Himebaugh, an 11-year-old child who seemingly vanished on Nov. 25, 1991. In the 30-plus years since Himebaugh went missing, law enforcement's strongest leads are a composite sketch of a person of interest and a theory that a convicted child sex predator, who's currently in prison, is involved. But neither are strong enough to bring charges or even advance the case.

Although the popular MNIST dataset \citep{mnist} is derived from the NIST database \citep{nist-sd19}, precise processing steps of this derivation have been lost to time. We propose a reconstruction that is accurate enough to serve as a replacement for the MNIST dataset, with insignificant changes in accuracy. We trace each MNIST digit to its NIST source and its rich metadata such as writer identifier, partition identifier, etc. We also reconstruct the complete MNIST test set with 60,000 samples instead of the usual 10,000. Since the balance 50,000 were never distributed, they enable us to investigate the impact of twenty-five years of MNIST experiments on the reported testing performances.

As a child of refugees, my parents' narrative is missing huge gaps of information. In our data rich world, archivists are finally piecing together new clues of history using unmanned systems to reopen cold cases. The Nazis were masters in using technology to mechanize killing and erasing all evidence of their crime. Nowhere is this more apparent than in Treblinka, Poland. The death camp exterminated close to 900,000 Jews over a 15-month period before a revolt led to its dismantlement in 1943.

Dutch Police are using artificial intelligence to crack unsolved cases, according to The Next Web. The national police force is working to digitize the more than 1,500 reports and 30 million pages of material in its cold case archive, only 15 percent of which is currently stored electronically. Once the transfer is complete, a machine learning algorithm will begin combing through the records and deciding which cases have the most promising evidence, reducing case processing time from weeks to a single day. "We're teaching the machine to do forensic screening," Jeroen Hammer, one of the architects of the system, told The Next Web. "The goal is that the AI can read cold cases we're currently digitizing, and decide which ones contain promising evidence that could lead to solving the case."

In August of 1944, Anne Frank and her family were captured by the Gestapo after spending a gruelling two years hidden in a secret annex within their apartment. The prolific diarist's work would posthumously bring her fame and recognition the world over. But, to this day, no one has been able to identify who was behind the betrayal that led to her death in a concentration camp. Fast forward 73 years, and a former FBI agent is betting artificial intelligence can help crack the mystery. Retired sleuth Vincent Pankoke, and his team of investigators (comprised of forensic scientists and members of the Dutch police force), are partnering with Amsterdam-based data company Xomnia on the ultimate cold case. As part of the newly-opened enquiry, a specially developed algorithm will scour reams of documents from the period.

In August of 1944, Anne Frank and her family were captured by the Gestapo after spending a gruelling two years hidden in a secret annex within their apartment. The prolific diarist's work would posthumously bring her fame and recognition the world over. But, to this day, no one has been able to identify who was behind the betrayal that led to her death in a concentration camp. Fast forward 73 years, and a former FBI agent is betting artificial intelligence can help crack the mystery. Retired sleuth Vincent Pankoke, and his team of investigators (comprised of forensic scientists and members of the Dutch police force), are partnering with Amsterdam-based data company Xomnia on the ultimate cold case. As part of the newly-opened enquiry, a specially developed algorithm will scour reams of documents from the period.

We first discuss certain problems with the classical probabilistic approach for assessing forensic evidence, in particular its inability to distinguish between lack of belief and disbelief, and its inability to model complete ignorance within a given population. We then discuss Shafer belief functions, a generalization of probability distributions, which can deal with both these objections. We use a calculus of belief functions which does not use the much criticized Dempster rule of combination, but only the very natural Dempster-Shafer conditioning. We then apply this calculus to some classical forensic problems like the various island problems and the problem of parental identification. If we impose no prior knowledge apart from assuming that the culprit or parent belongs to a given population (something which is possible in our setting), then our answers differ from the classical ones when uniform or other priors are imposed. We can actually retrieve the classical answers by imposing the relevant priors, so our setup can and should be interpreted as a generalization of the classical methodology, allowing more flexibility. We show how our calculus can be used to develop an analogue of Bayes' rule, with belief functions instead of classical probabilities. We also discuss consequences of our theory for legal practice.