Tag Archives: machine learning algorithms

Being smart about using artificial intelligence in the field of medicine

Since my August 20, 2018 post featured an opinion piece about the possibly imminent replacement of radiologists with artificial intelligence systems and the latest research about employing them for diagnosing eye diseases, it seems like a good time to examine some of the mythology embedded in the discussion about AI and medicine.

Imperfections in medical AI systems

An August 15, 2018 article for Slate.com by W. Nicholson Price II (who teaches at the University of Michigan School of Law; in addition to his law degree he has a PhD in Biological Sciences from Columbia University) begins with the peppy, optimistic view before veering into more critical territory (Note: Links have been removed),

For millions of people suffering from diabetes, new technology enabled by artificial intelligence promises to make management much easier. Medtronic’s Guardian Connect system promises to alert users 10 to 60 minutes before they hit high or low blood sugar level thresholds, thanks to IBM Watson, “the same supercomputer technology that can predict global weather patterns.” Startup Beta Bionics goes even further: In May, it received Food and Drug Administration approval to start clinical trials on what it calls a “bionic pancreas system” powered by artificial intelligence, capable of “automatically and autonomously managing blood sugar levels 24/7.”

An artificial pancreas powered by artificial intelligence represents a huge step forward for the treatment of diabetes—but getting it right will be hard. Artificial intelligence (also known in various iterations as deep learning and machine learning) promises to automatically learn from patterns in medical data to help us do everything from managing diabetes to finding tumors in an MRI to predicting how long patients will live. But the artificial intelligence techniques involved are typically opaque. We often don’t know how the algorithm makes the eventual decision. And they may change and learn from new data—indeed, that’s a big part of the promise. But when the technology is complicated, opaque, changing, and absolutely vital to the health of a patient, how do we make sure it works as promised?

Price describes how a ‘closed loop’ artificial pancreas with AI would automate insulin levels for diabetic patients, flaws in the automated system, and how companies like to maintain a competitive advantage (Note: Links have been removed),

[…] a “closed loop” artificial pancreas, where software handles the whole issue, receiving and interpreting signals from the monitor, deciding when and how much insulin is needed, and directing the insulin pump to provide the right amount. The first closed-loop system was approved in late 2016. The system should take as much of the issue off the mind of the patient as possible (though, of course, that has limits). Running a close-loop artificial pancreas is challenging. The way people respond to changing levels of carbohydrates is complicated, as is their response to insulin; it’s hard to model accurately. Making it even more complicated, each individual’s body reacts a little differently.

Here’s where artificial intelligence comes into play. Rather than trying explicitly to figure out the exact model for how bodies react to insulin and to carbohydrates, machine learning methods, given a lot of data, can find patterns and make predictions. And existing continuous glucose monitors (and insulin pumps) are excellent at generating a lot of data. The idea is to train artificial intelligence algorithms on vast amounts of data from diabetic patients, and to use the resulting trained algorithms to run a closed-loop artificial pancreas. Even more exciting, because the system will keep measuring blood glucose, it can learn from the new data and each patient’s artificial pancreas can customize itself over time as it acquires new data from that patient’s particular reactions.

Here’s the tough question: How will we know how well the system works? Diabetes software doesn’t exactly have the best track record when it comes to accuracy. A 2015 study found that among smartphone apps for calculating insulin doses, two-thirds of the apps risked giving incorrect results, often substantially so. … And companies like to keep their algorithms proprietary for a competitive advantage, which makes it hard to know how they work and what flaws might have gone unnoticed in the development process.

There’s more,

These issues aren’t unique to diabetes care—other A.I. algorithms will also be complicated, opaque, and maybe kept secret by their developers. The potential for problems multiplies when an algorithm is learning from data from an entire hospital, or hospital system, or the collected data from an entire state or nation, not just a single patient. …

The [US Food and Drug Administraiont] FDA is working on this problem. The head of the agency has expressed his enthusiasm for bringing A.I. safely into medical practice, and the agency has a new Digital Health Innovation Action Plan to try to tackle some of these issues. But they’re not easy, and one thing making it harder is a general desire to keep the algorithmic sauce secret. The example of IBM Watson for Oncology has given the field a bit of a recent black eye—it turns out that the company knew the algorithm gave poor recommendations for cancer treatment but kept that secret for more than a year. …

While Price focuses on problems with algorithms and with developers and their business interests, he also hints at some of the body’s complexities.

Can AI systems be like people?

Susan Baxter, a medical writer with over 20 years experience, a PhD in health economics, and author of countless magazine articles and several books, offers a more person-centered approach to the discussion in her July 6, 2018 posting on susanbaxter.com,

The fascination with AI continues to irk, given that every second thing I read seems to be extolling the magic of AI and medicine and how It Will Change Everything. Which it will not, trust me. The essential issue of illness remains perennial and revolves around an individual for whom no amount of technology will solve anything without human contact. …

But in this world, or so we are told by AI proponents, radiologists will soon be obsolete. [my August 20, 2018 post] The adaptational learning capacities of AI mean that reading a scan or x-ray will soon be more ably done by machines than humans. The presupposition here is that we, the original programmers of this artificial intelligence, understand the vagaries of real life (and real disease) so wonderfully that we can deconstruct these much as we do the game of chess (where, let’s face it, Big Blue ate our lunch) and that analyzing a two-dimensional image of a three-dimensional body, already problematic, can be reduced to a series of algorithms.

Attempting to extrapolate what some “shadow” on a scan might mean in a flesh and blood human isn’t really quite the same as bishop to knight seven. Never mind the false positive/negatives that are considered an acceptable risk or the very real human misery they create.

Moravec called it

It’s called Moravec’s paradox, the inability of humans to realize just how complex basic physical tasks are – and the corresponding inability of AI to mimic it. As you walk across the room, carrying a glass of water, talking to your spouse/friend/cat/child; place the glass on the counter and open the dishwasher door with your foot as you open a jar of pickles at the same time, take a moment to consider just how many concurrent tasks you are doing and just how enormous the computational power these ostensibly simple moves would require.

Researchers in Singapore taught industrial robots to assemble an Ikea chair. Essentially, screw in the legs. A person could probably do this in a minute. Maybe two. The preprogrammed robots took nearly half an hour. And I suspect programming those robots took considerably longer than that.

Ironically, even Elon Musk, who has had major production problems with the Tesla cars rolling out of his high tech factory, has conceded (in a tweet) that “Humans are underrated.”

I wouldn’t necessarily go that far given the political shenanigans of Trump & Co. but in the grand scheme of things I tend to agree. …

Is AI going the way of gene therapy?

Susan draws a parallel between the AI and medicine discussion with the discussion about genetics and medicine (Note: Links have been removed),

On a somewhat similar note – given the extent to which genetics discourse has that same linear, mechanistic  tone [as AI and medicine] – it turns out all this fine talk of using genetics to determine health risk and whatnot is based on nothing more than clever marketing, since a lot of companies are making a lot of money off our belief in DNA. Truth is half the time we don’t even know what a gene is never mind what it actually does;  geneticists still can’t agree on how many genes there are in a human genome, as this article in Nature points out.

Along the same lines, I was most amused to read about something called the Super Seniors Study, research following a group of individuals in their 80’s, 90’s and 100’s who seem to be doing really well. Launched in 2002 and headed by Angela Brooks Wilson, a geneticist at the BC [British Columbia] Cancer Agency and SFU [Simon Fraser University] Chair of biomedical physiology and kinesiology, this longitudinal work is examining possible factors involved in healthy ageing.

Turns out genes had nothing to do with it, the title of the Globe and Mail article notwithstanding. (“Could the DNA of these super seniors hold the secret to healthy aging?” The answer, a resounding “no”, well hidden at the very [end], the part most people wouldn’t even get to.) All of these individuals who were racing about exercising and working part time and living the kind of life that makes one tired just reading about it all had the same “multiple (genetic) factors linked to a high probability of disease”. You know, the gene markers they tell us are “linked” to cancer, heart disease, etc., etc. But these super seniors had all those markers but none of the diseases, demonstrating (pretty strongly) that the so-called genetic links to disease are a load of bunkum. Which (she said modestly) I have been saying for more years than I care to remember. You’re welcome.

The fundamental error in this type of linear thinking is in allowing our metaphors (genes are the “blueprint” of life) and propensity towards social ideas of determinism to overtake common sense. Biological and physiological systems are not static; they respond to and change to life in its entirety, whether it’s diet and nutrition to toxic or traumatic insults. Immunity alters, endocrinology changes, – even how we think and feel affects the efficiency and effectiveness of physiology. Which explains why as we age we become increasingly dissimilar.

If you have the time, I encourage to read Susan’s comments in their entirety.

Scientific certainties

Following on with genetics, gene therapy dreams, and the complexity of biology, the June 19, 2018 Nature article by Cassandra Willyard (mentioned in Susan’s posting) highlights an aspect of scientific research not often mentioned in public,

One of the earliest attempts to estimate the number of genes in the human genome involved tipsy geneticists, a bar in Cold Spring Harbor, New York, and pure guesswork.

That was in 2000, when a draft human genome sequence was still in the works; geneticists were running a sweepstake on how many genes humans have, and wagers ranged from tens of thousands to hundreds of thousands. Almost two decades later, scientists armed with real data still can’t agree on the number — a knowledge gap that they say hampers efforts to spot disease-related mutations.

In 2000, with the genomics community abuzz over the question of how many human genes would be found, Ewan Birney launched the GeneSweep contest. Birney, now co-director of the European Bioinformatics Institute (EBI) in Hinxton, UK, took the first bets at a bar during an annual genetics meeting, and the contest eventually attracted more than 1,000 entries and a US$3,000 jackpot. Bets on the number of genes ranged from more than 312,000 to just under 26,000, with an average of around 40,000. These days, the span of estimates has shrunk — with most now between 19,000 and 22,000 — but there is still disagreement (See ‘Gene Tally’).

… the inconsistencies in the number of genes from database to database are problematic for researchers, Pruitt says. “People want one answer,” she [Kim Pruitt, a genome researcher at the US National Center for Biotechnology Information {NCB}] in Bethesda, Maryland] adds, “but biology is complex.”

I wanted to note that scientists do make guesses and not just with genetics. For example, Gina Mallet’s 2005 book ‘Last Chance to Eat: The Fate of Taste in a Fast Food World’ recounts the story of how good and bad levels of cholesterol were established—the experts made some guesses based on their experience. That said, Willyard’s article details the continuing effort to nail down the number of genes almost 20 years after the human genome project was completed and delves into the problems the scientists have uncovered.

Final comments

In addition to opaque processes with developers/entrepreneurs wanting to maintain their secrets for competitive advantages and in addition to our own poor understanding of the human body (how many genes are there anyway?), there are same major gaps (reflected in AI) in our understanding of various diseases. Angela Lashbrook’s August 16, 2018 article for The Atlantic highlights some issues with skin cancer and shade of your skin (Note: Links have been removed),

… While fair-skinned people are at the highest risk for contracting skin cancer, the mortality rate for African Americans is considerably higher: Their five-year survival rate is 73 percent, compared with 90 percent for white Americans, according to the American Academy of Dermatology.

As the rates of melanoma for all Americans continue a 30-year climb, dermatologists have begun exploring new technologies to try to reverse this deadly trend—including artificial intelligence. There’s been a growing hope in the field that using machine-learning algorithms to diagnose skin cancers and other skin issues could make for more efficient doctor visits and increased, reliable diagnoses. The earliest results are promising—but also potentially dangerous for darker-skinned patients.

… Avery Smith, … a software engineer in Baltimore, Maryland, co-authored a paper in JAMA [Journal of the American Medical Association] Dermatology that warns of the potential racial disparities that could come from relying on machine learning for skin-cancer screenings. Smith’s co-author, Adewole Adamson of the University of Texas at Austin, has conducted multiple studies on demographic imbalances in dermatology. “African Americans have the highest mortality rate [for skin cancer], and doctors aren’t trained on that particular skin type,” Smith told me over the phone. “When I came across the machine-learning software, one of the first things I thought was how it will perform on black people.”

Recently, a study that tested machine-learning software in dermatology, conducted by a group of researchers primarily out of Germany, found that “deep-learning convolutional neural networks,” or CNN, detected potentially cancerous skin lesions better than the 58 dermatologists included in the study group. The data used for the study come from the International Skin Imaging Collaboration, or ISIC, an open-source repository of skin images to be used by machine-learning algorithms. Given the rise in melanoma cases in the United States, a machine-learning algorithm that assists dermatologists in diagnosing skin cancer earlier could conceivably save thousands of lives each year.

… Chief among the prohibitive issues, according to Smith and Adamson, is that the data the CNN relies on come from primarily fair-skinned populations in the United States, Australia, and Europe. If the algorithm is basing most of its knowledge on how skin lesions appear on fair skin, then theoretically, lesions on patients of color are less likely to be diagnosed. “If you don’t teach the algorithm with a diverse set of images, then that algorithm won’t work out in the public that is diverse,” says Adamson. “So there’s risk, then, for people with skin of color to fall through the cracks.”

As Adamson and Smith’s paper points out, racial disparities in artificial intelligence and machine learning are not a new issue. Algorithms have mistaken images of black people for gorillas, misunderstood Asians to be blinking when they weren’t, and “judged” only white people to be attractive. An even more dangerous issue, according to the paper, is that decades of clinical research have focused primarily on people with light skin, leaving out marginalized communities whose symptoms may present differently.

The reasons for this exclusion are complex. According to Andrew Alexis, a dermatologist at Mount Sinai, in New York City, and the director of the Skin of Color Center, compounding factors include a lack of medical professionals from marginalized communities, inadequate information about those communities, and socioeconomic barriers to participating in research. “In the absence of a diverse study population that reflects that of the U.S. population, potential safety or efficacy considerations could be missed,” he says.

Adamson agrees, elaborating that with inadequate data, machine learning could misdiagnose people of color with nonexistent skin cancers—or miss them entirely. But he understands why the field of dermatology would surge ahead without demographically complete data. “Part of the problem is that people are in such a rush. This happens with any new tech, whether it’s a new drug or test. Folks see how it can be useful and they go full steam ahead without thinking of potential clinical consequences. …

Improving machine-learning algorithms is far from the only method to ensure that people with darker skin tones are protected against the sun and receive diagnoses earlier, when many cancers are more survivable. According to the Skin Cancer Foundation, 63 percent of African Americans don’t wear sunscreen; both they and many dermatologists are more likely to delay diagnosis and treatment because of the belief that dark skin is adequate protection from the sun’s harmful rays. And due to racial disparities in access to health care in America, African Americans are less likely to get treatment in time.

Happy endings

I’ll add one thing to Price’s article, Susan’s posting, and Lashbrook’s article about the issues with AI , certainty, gene therapy, and medicine—the desire for a happy ending prefaced with an easy solution. If the easy solution isn’t possible accommodations will be made but that happy ending is a must. All disease will disappear and there will be peace on earth. (Nod to Susan Baxter and her many discussions with me about disease processes and happy endings.)

The solutions, for the most part, are seen as technological despite the mountain of evidence suggesting that technology reflects our own imperfect understanding of health and disease therefore providing what is at best an imperfect solution.

Also, we tend to underestimate just how complex humans are not only in terms of disease and health but also with regard to our skills, understanding, and, perhaps not often enough, our ability to respond appropriately in the moment.

There is much to celebrate in what has been accomplished: no more black death, no more smallpox, hip replacements, pacemakers, organ transplants, and much more. Yes, we should try to improve our medicine. But, maybe alongside the celebration we can welcome AI and other technologies with a lot less hype and a lot more skepticism.

Changing synaptic connectivity with a memristor

The French have announced some research into memristive devices that mimic both short-term and long-term neural plasticity according to a Dec. 6, 2016 news item on Nanowerk,

Leti researchers have demonstrated that memristive devices are excellent candidates to emulate synaptic plasticity, the capability of synapses to enhance or diminish their connectivity between neurons, which is widely believed to be the cellular basis for learning and memory.

The breakthrough was presented today [Dec. 6, 2016] at IEDM [International Electron Devices Meeting] 2016 in San Francisco in the paper, “Experimental Demonstration of Short and Long Term Synaptic Plasticity Using OxRAM Multi k-bit Arrays for Reliable Detection in Highly Noisy Input Data”.

Neural systems such as the human brain exhibit various types and time periods of plasticity, e.g. synaptic modifications can last anywhere from seconds to days or months. However, prior research in utilizing synaptic plasticity using memristive devices relied primarily on simplified rules for plasticity and learning.

The project team, which includes researchers from Leti’s sister institute at CEA Tech, List, along with INSERM and Clinatec, proposed an architecture that implements both short- and long-term plasticity (STP and LTP) using RRAM devices.

A Dec. 6, 2016 Laboratoire d’électronique des technologies de l’information (LETI) press release, which originated the news item, elaborates,

“While implementing a learning rule for permanent modifications – LTP, based on spike-timing-dependent plasticity – we also incorporated the possibility of short-term modifications with STP, based on the Tsodyks/Markram model,” said Elisa Vianello, Leti non-volatile memories and cognitive computing specialist/research engineer. “We showed the benefits of utilizing both kinds of plasticity with visual pattern extraction and decoding of neural signals. LTP allows our artificial neural networks to learn patterns, and STP makes the learning process very robust against environmental noise.”

Resistive random-access memory (RRAM) devices coupled with a spike-coding scheme are key to implementing unsupervised learning with minimal hardware footprint and low power consumption. Embedding neuromorphic learning into low-power devices could enable design of autonomous systems, such as a brain-machine interface that makes decisions based on real-time, on-line processing of in-vivo recorded biological signals. Biological data are intrinsically highly noisy and the proposed combined LTP and STP learning rule is a powerful technique to improve the detection/recognition rate. This approach may enable the design of autonomous implantable devices for rehabilitation purposes

Leti, which has worked on RRAM to develop hardware neuromorphic architectures since 2010, is the coordinator of the H2020 [Horizon 2020] European project NeuRAM3. That project is working on fabricating a chip with architecture that supports state-of-the-art machine-learning algorithms and spike-based learning mechanisms.

That’s it folks.

Removing gender-based stereotypes from algorithms

Most people don’t think of algorithms as having biases and stereotypes but Michael Zou in his Sept. 26, 2016 essay for The Conversation (h/t phys.org Sept. 26, 2016 news item) says different, Note: Links have been removed,

Machine learning is ubiquitous in our daily lives. Every time we talk to our smartphones, search for images or ask for restaurant recommendations, we are interacting with machine learning algorithms. They take as input large amounts of raw data, like the entire text of an encyclopedia, or the entire archives of a newspaper, and analyze the information to extract patterns that might not be visible to human analysts. But when these large data sets include social bias, the machines learn that too.

A machine learning algorithm is like a newborn baby that has been given millions of books to read without being taught the alphabet or knowing any words or grammar. The power of this type of information processing is impressive, but there is a problem. When it takes in the text data, a computer observes relationships between words based on various factors, including how often they are used together.

We can test how well the word relationships are identified by using analogy puzzles. Suppose I ask the system to complete the analogy “He is to King as She is to X.” If the system comes back with “Queen,” then we would say it is successful, because it returns the same answer a human would.

Our research group trained the system on Google News articles, and then asked it to complete a different analogy: “Man is to Computer Programmer as Woman is to X.” The answer came back: “Homemaker.”

Zou explains how a machine (algorithm) learns and then notes this,

Not only can the algorithm reflect society’s biases – demonstrating how much those biases are contained in the input data – but the system can potentially amplify gender stereotypes. Suppose I search for “computer programmer” and the search program uses a gender-biased database that associates that term more closely with a man than a woman.

The search results could come back flawed by the bias. Because “John” as a male name is more closely related to “computer programmer” than the female name “Mary” in the biased data set, the search program could evaluate John’s website as more relevant to the search than Mary’s – even if the two websites are identical except for the names and gender pronouns.

It’s true that the biased data set could actually reflect factual reality – perhaps there are more “Johns” who are programmers than there are “Marys” – and the algorithms simply capture these biases. This does not absolve the responsibility of machine learning in combating potentially harmful stereotypes. The biased results would not just repeat but could even boost the statistical bias that most programmers are male, by moving the few female programmers lower in the search results. It’s useful and important to have an alternative that’s not biased.

There is a way according to Zou that stereotypes can be removed,

Our debiasing system uses real people to identify examples of the types of connections that are appropriate (brother/sister, king/queen) and those that should be removed. Then, using these human-generated distinctions, we quantified the degree to which gender was a factor in those word choices – as opposed to, say, family relationships or words relating to royalty.

Next we told our machine-learning algorithm to remove the gender factor from the connections in the embedding. This removes the biased stereotypes without reducing the overall usefulness of the embedding.

When that is done, we found that the machine learning algorithm no longer exhibits blatant gender stereotypes. We are investigating applying related ideas to remove other types of biases in the embedding, such as racial or cultural stereotypes.

If you have time, I encourage you to read the essay in its entirety and this June 14, 2016 posting about research into algorithms and how they make decisions for you about credit, medical diagnoses, job opportunities and more.

There’s also an Oct. 24, 2016 article by Michael Light on Salon.com on the topic (Note: Links have been removed),

In a recent book that was longlisted for the National Book Award, Cathy O’Neil, a data scientist, blogger and former hedge-fund quant, details a number of flawed algorithms to which we have given incredible power — she calls them “Weapons of Math Destruction.” We have entrusted these WMDs to make important, potentially life-altering decisions, yet in many cases, they embed human race and class biases; in other cases, they don’t function at all.
Among other examples, O’Neil examines a “value-added” model New York City used to decide which teachers to fire, even though, she writes, the algorithm was useless, functioning essentially as a random number generator, arbitrarily ending careers. She looks at models put to use by judges to assign recidivism scores to inmates that ended up having a racist inclination. And she looks at how algorithms are contributing to American partisanship, allowing political operatives to target voters with information that plays to their existing biases and fears.

I recommend reading Light’s article in its entirety.

Accountability for artificial intelligence decision-making

How does an artificial intelligence program arrive at its decisions? It’s a question that’s not academic any more as these programs take on more decision-making chores according to a May 25, 2016 Carnegie Mellon University news release (also on EurekAlert) by Bryon Spice (Note: Links have been removed),

Machine-learning algorithms increasingly make decisions about credit, medical diagnoses, personalized recommendations, advertising and job opportunities, among other things, but exactly how usually remains a mystery. Now, new measurement methods developed by Carnegie Mellon University [CMU] researchers could provide important insights to this process.

Was it a person’s age, gender or education level that had the most influence on a decision? Was it a particular combination of factors? CMU’s Quantitative Input Influence (QII) measures can provide the relative weight of each factor in the final decision, said Anupam Datta, associate professor of computer science and electrical and computer engineering.

It’s reassuring to know that more requests for transparency of the decision-making process are being made. After all, it’s disconcerting that someone with the life experience of a gnat and/or possibly some issues might be developing an algorithm that could affection your life in some fundamental ways. Here’s more from the news release (Note: Links have been removed),

“Demands for algorithmic transparency are increasing as the use of algorithmic decision-making systems grows and as people realize the potential of these systems to introduce or perpetuate racial or sex discrimination or other social harms,” Datta said.

“Some companies are already beginning to provide transparency reports, but work on the computational foundations for these reports has been limited,” he continued. “Our goal was to develop measures of the degree of influence of each factor considered by a system, which could be used to generate transparency reports.”

These reports might be generated in response to a particular incident — why an individual’s loan application was rejected, or why police targeted an individual for scrutiny, or what prompted a particular medical diagnosis or treatment. Or they might be used proactively by an organization to see if an artificial intelligence system is working as desired, or by a regulatory agency to see whether a decision-making system inappropriately discriminated between groups of people.

Datta, along with Shayak Sen, a Ph.D. student in computer science, and Yair Zick, a post-doctoral researcher in the Computer Science Department, will present their report on QII at the IEEE Symposium on Security and Privacy, May 23–25 [2016], in San Jose, Calif.

Generating these QII measures requires access to the system, but doesn’t necessitate analyzing the code or other inner workings of the system, Datta said. It also requires some knowledge of the input dataset that was initially used to train the machine-learning system.

A distinctive feature of QII measures is that they can explain decisions of a large class of existing machine-learning systems. A significant body of prior work takes a complementary approach, redesigning machine-learning systems to make their decisions more interpretable and sometimes losing prediction accuracy in the process.

QII measures carefully account for correlated inputs while measuring influence. For example, consider a system that assists in hiring decisions for a moving company. Two inputs, gender and the ability to lift heavy weights, are positively correlated with each other and with hiring decisions. Yet transparency into whether the system uses weight-lifting ability or gender in making its decisions has substantive implications for determining if it is engaging in discrimination.

“That’s why we incorporate ideas for causal measurement in defining QII,” Sen said. “Roughly, to measure the influence of gender for a specific individual in the example above, we keep the weight-lifting ability fixed, vary gender and check whether there is a difference in the decision.”

Observing that single inputs may not always have high influence, the QII measures also quantify the joint influence of a set of inputs, such as age and income, on outcomes and the marginal influence of each input within the set. Since a single input may be part of multiple influential sets, the average marginal influence of the input is computed using principled game-theoretic aggregation measures previously applied to measure influence in revenue division and voting.

“To get a sense of these influence measures, consider the U.S. presidential election,” Zick said. “California and Texas have influence because they have many voters, whereas Pennsylvania and Ohio have power because they are often swing states. The influence aggregation measures we employ account for both kinds of power.”

The researchers tested their approach against some standard machine-learning algorithms that they used to train decision-making systems on real data sets. They found that the QII provided better explanations than standard associative measures for a host of scenarios they considered, including sample applications for predictive policing and income prediction.

Now, they are seeking collaboration with industrial partners so that they can employ QII at scale on operational machine-learning systems.

Here’s a link to and a citation for a PDF of the paper presented at the May 2016 conference,

Algorithmic Transparency via Quantitative Input Influence: Theory and Experiments with Learning Systems by Anupam Datta, Shayak Sen, Yair Zick. Presented at the at the IEEE Symposium on Security and Privacy, May 23–25, in San Jose, Calif.

I’ve also embedded the paper here,

CarnegieMellon_AlgorithmicTransparency

Performances Tom Hanks never gave

The answer to the question, “What makes Tom Hanks look like Tom  Hanks?” leads to machine learning and algorithms according to a Dec. 7, 2015 University of Washington University news release (also on EurekAlert) Note: Link have been removed,

Tom Hanks has appeared in many acting roles over the years, playing young and old, smart and simple. Yet we always recognize him as Tom Hanks.

Why? Is it his appearance? His mannerisms? The way he moves?

University of Washington researchers have demonstrated that it’s possible for machine learning algorithms to capture the “persona” and create a digital model of a well-photographed person like Tom Hanks from the vast number of images of them available on the Internet.

With enough visual data to mine, the algorithms can also animate the digital model of Tom Hanks to deliver speeches that the real actor never performed.

“One answer to what makes Tom Hanks look like Tom Hanks can be demonstrated with a computer system that imitates what Tom Hanks will do,” said lead author Supasorn Suwajanakorn, a UW graduate student in computer science and engineering.

As for the performances Tom Hanks never gave, the news release offers more detail,

The technology relies on advances in 3-D face reconstruction, tracking, alignment, multi-texture modeling and puppeteering that have been developed over the last five years by a research group led by UW assistant professor of computer science and engineering Ira Kemelmacher-Shlizerman. The new results will be presented in a paper at the International Conference on Computer Vision in Chile on Dec. 16.

The team’s latest advances include the ability to transfer expressions and the way a particular person speaks onto the face of someone else — for instance, mapping former president George W. Bush’s mannerisms onto the faces of other politicians and celebrities.

Here’s a video demonstrating how former President Bush’s speech and mannerisms have mapped onto other famous faces including Hanks’s,

The research team has future plans for this technology (from the news release)

It’s one step toward a grand goal shared by the UW computer vision researchers: creating fully interactive, three-dimensional digital personas from family photo albums and videos, historic collections or other existing visuals.

As virtual and augmented reality technologies develop, they envision using family photographs and videos to create an interactive model of a relative living overseas or a far-away grandparent, rather than simply Skyping in two dimensions.

“You might one day be able to put on a pair of augmented reality glasses and there is a 3-D model of your mother on the couch,” said senior author Kemelmacher-Shlizerman. “Such technology doesn’t exist yet — the display technology is moving forward really fast — but how do you actually re-create your mother in three dimensions?”

One day the reconstruction technology could be taken a step further, researchers say.

“Imagine being able to have a conversation with anyone you can’t actually get to meet in person — LeBron James, Barack Obama, Charlie Chaplin — and interact with them,” said co-author Steve Seitz, UW professor of computer science and engineering. “We’re trying to get there through a series of research steps. One of the true tests is can you have them say things that they didn’t say but it still feels like them? This paper is demonstrating that ability.”

Existing technologies to create detailed three-dimensional holograms or digital movie characters like Benjamin Button often rely on bringing a person into an elaborate studio. They painstakingly capture every angle of the person and the way they move — something that can’t be done in a living room.

Other approaches still require a person to be scanned by a camera to create basic avatars for video games or other virtual environments. But the UW computer vision experts wanted to digitally reconstruct a person based solely on a random collection of existing images.

To reconstruct celebrities like Tom Hanks, Barack Obama and Daniel Craig, the machine learning algorithms mined a minimum of 200 Internet images taken over time in various scenarios and poses — a process known as learning ‘in the wild.’

“We asked, ‘Can you take Internet photos or your personal photo collection and animate a model without having that person interact with a camera?'” said Kemelmacher-Shlizerman. “Over the years we created algorithms that work with this kind of unconstrained data, which is a big deal.”

Suwajanakorn more recently developed techniques to capture expression-dependent textures — small differences that occur when a person smiles or looks puzzled or moves his or her mouth, for example.

By manipulating the lighting conditions across different photographs, he developed a new approach to densely map the differences from one person’s features and expressions onto another person’s face. That breakthrough enables the team to ‘control’ the digital model with a video of another person, and could potentially enable a host of new animation and virtual reality applications.

“How do you map one person’s performance onto someone else’s face without losing their identity?” said Seitz. “That’s one of the more interesting aspects of this work. We’ve shown you can have George Bush’s expressions and mouth and movements, but it still looks like George Clooney.”

Here’s a link to and a citation for the paper presented at the conference in Chile,

What Makes Tom Hanks Look Like Tom Hanks by Supasorn Suwajanakorn, Steven M. Seitz, Ira Kemelmacher-Shlizerman for the 2015 ICCV conference, Dec. 13 – 15, 2015 in Chile.

You can find out more about the conference here.