Category: Blogs

Posted on

Knowing the unknown by digging deep

Kishore Jethanandani

Deep learning, referred to as neural network algorithms, is a lot like solving a crossword puzzle–the unknowns in gargantuan data stores are knowable only by their relationships with the known. Unsupervised deep learning goes further and does not presume, at the outset, any knowledge of the interdependencies in the data.

Supervised deep learning is analogous to searching for an undersea destination like an oil well with the knowledge of the coastline alone. It reads the known relationships in the geophysical data in the layers underneath the seashore to reach, progressively, the oil well. Unsupervised learning first establishes whether a relationship exists between the contours of the coastline and the subterranean topography.

We spoke to Dr. Charles H Martin, a long-time expert in machine learning and the founder of Calculation Consulting, about the prospects for enterprise applications of supervised and unsupervised deep learning. “Many in the business world recognize the vast potential of applications of deep learning and the technology has matured for widespread adoption,” Dr. Martin surmised. “The most hospitable culture for machine learning is scientific and open to recurring experimentation with ideas and evolving business models, the legacy enterprise fixation on engineering and static processes is a barrier to its progress,” Dr. Martin underscored.

Unstructured data abounds, and the familiar methods of analyzing them with categories and correlations do not necessarily exist. The size and variety of such databases can elude modeling. These unstructured databases have valuable information like social media conversations about brands, video from traffic cameras, sensor data of factory equipment, or trading data from exchanges that are akin to finding a needle in a haystack. Deep learning algorithms find the brand value from positive and negative remarks on social media, elusive fugitives in the video from traffic cameras, the failing equipment from the factory data, or the investment opportunity in the trading data.

“Unsupervised deep learning helps in detecting patterns and hypothesis formulation while supervised deep learning is for hypothesis testing and deeper exploration,” Dr. Martin concluded.  “Unsupervised deep learning has proved to be useful for fraud detection, and oil exploration—anomalies in the data point to cybercrime and oil respectively,” he explained. “The prediction of corporate performance using granular data such as satellite imagery of traffic in the parking lots of retail companies is an example of the second generation of supervised deep learning,” Dr. Martin revealed.

Early detection of illnesses from medical imaging is one category of problems that deep learning is well suited to address. Citing the example of COPD (Chronic Obstructive Pulmonary Disease), Dave Sullivan, the CEO of Ersatz Labs, a cloud-based deep learning company based in San Francisco, told us, “the imaging data shows nodules and not all of them indicate COPD. It is hard for even a trained eye to tell one from another. Deep learning techniques evolve as they are calibrated and recalibrated (trained) on vast volumes of data gathered in the past, and they learn to distinguish with a high degree of accuracy for individual cases.”

Clarifai has democratized access to its deep learning with its API, which allows holders of data to analyze and benefit from the insights.  We spoke to Matthew Zeiler, the CEO and Founder of Clarifai, to understand how its partners use the technology.  One of them is France-based i-nside.com, a healthcare company, which employs smartphones to conduct routine examinations of the mouth, ear, and throat to generate data for diagnosis. “In developing countries where doctors are scarce, the analysis of the data points to therapies that are reliable,” Zeiler told us. “In developed countries, the analysis of the data supports the judgment of doctors, and they have reported satisfactory results,” Zeiler added.

Enterprise is not the only place where deep learning has found a home—consumer applications like Google Now, Microsoft’s Cortana and Assistant are available in the market. Folks are often anxious and distracted, at work or play, when they are unable to keep track of critical events that could affect them or their family. Home surveillance watches pets, the return of young children from school, elderly relatives falling, the arrival of critical packages and more. What matters is an alert on an unusual event. Camio uses the camera of a handheld phone or any other home device like a computer to capture video of happenings at home. When something irregular happens, IFTTT sends alerts.

Deep learning mimics the neurons of the brain to sift the meaningful relationships otherwise lost in the clutter of humongous streams or data stores. Machines can do it faster when the correlations are known. When they are unknown, it helps to discern the patterns before deciding to invest time in deeper investigations.

 

Posted on

Indiana Jones in the age of analytics

By Kishore Jethanandani

Drilling engineers navigate hazardous oil wells in earth’s dark hollows not in the manner of the swashbuckling Indiana Jones but collaboratively with staid geophysicists and geologists who parse terabytes of data to calculate the risk of the next cascade of rocks or an explosion. 3D visuals of seismological data, superimposed with sensor fed real-time data, help offsite professionals to collaborate with engineers working at well-sites.

The drilling machines also generate streams of data with an array of sensors used for well logging. The data is transmitted to remote sites where it is aggregated and is accessible to geologists and geophysicists.  These sensors can read geological data such as hydrocarbon bearing capacity of rocks as well as the data related to the rig operations such as the borehole pressure, vibrations, weight on the drill and its direction and much more all in the context of the well environment.

A major breakthrough has been achieved with the ability to pool data coming from a variety of sources in a single repository with help from standards under the rubric of Wellsite Information Transfer Specification (WITS).  The availability of a storehouse of well data opens the way for a bevy of firms, specialized in reading the geological and operations data, to find patterns and guide engineers to find the most optimal ways to explore and produce oil in complex wells in the depths of the earth and the oceans.

A typical case of oil and exploration companies encountering Catch-22 situations is that of PEMEX which ran into the dead end of a salt dome underground. The alternative was to circumvent the dome.  The rub was the risk of getting into a quagmire of mud. A game plan was crafted over eight months in collaboration with a multidisciplinary staff that included preliminary testing, 3D modeling and simulation and contingency planning. The entire exercise determined that exceptional pressures were likely to be encountered due to the presence of the salt dome in the vicinity of the alternative route compounded by a host of other probable risks.

For managing the risks, a predictive model was written based on the available geological data and its performance was monitored by comparing it with the actual performance data, generated during drilling. The variance between the predicted values and actuals revealed unanticipated hazards and informed action plans that engineers could use to deal with risks.

The oilfields today range from the icy Artic with its shifting icebergs and thawing permafrost to the raging storms at the deeper ends of oceans with their easily ignited submarine methane and the cavernous rocks of shale oil. Oil companies look to protect their multi-billion dollar investments in the projects and their staff from certain death if any of the risks are misjudged.

Fortunately, oil companies have accumulated multidimensional data now available on a standard platform. 3D video collaboration brings together the human talent from distant locations to crack the codes that help to improve the standards of safety and effectiveness.

A variant of this post was published in the now defunct Collaborative Planet hosted by UBM Techweb

Posted on

Future of Healthcare in the USA: how it could be the growth engine

By Kishore Jethanandani

Kondratieff cycles, which span thirty to fifty years, are marked by breakthroughs in technology and reform of institutions that drive expansion and a downturn sets in as technologies mature and unnoticed dysfunction in institutions surfaces. This was true for information technologies over the last cycle. Computers were a curiosity as long as the use of behemoth mainframes was confined to some large enterprises. The turning point was the introduction of mini-computers and then personal computers in the 1980s which became, with the advent of the Windows user interface, as much a part of everyday life as the washing machine and the refrigerator over the 1990s.

Concurrently, regulatory reform of the telecommunications industry dismantled the AT&T monopoly; the costs of communications plummeted and paved the way for a pervasive Internet. Business solutions became the buzzword of the last decade as information technologies found new uses in a broad range of industries. For the consumer, the home computer and the associated software became as essential as the furniture in their home. The mobile Internet has growing number of applications that make the wireless phone as indispensable as a wallet. The momentum in the growth of wireless phones is expected to be maintained till 2014 partially offsetting the slowdown in the rest of the information technology industry.

The high growth rates that were propelled by innovation in information technologies in the 1990s began to slow down over the last decade. A recent study completed by well-known economist, Dale W Jorgenson of the Harvard University and his colleagues, confirms that growth rates decelerated over the last decade. The average growth of value added in the industrial economy over the period 1960 to 2007 was 3.45%, with a peak of 4.52% between 1995 and 2000, which slowed down to 2.78% between 2000 and 2007. The corresponding growth of the IT producing industries was an average of 15.92% over 1960 and 2007, peaked at 27.35% between 1995 and 2000, and slowed down to 10.19% between 2000 and 2007. IT using industries had an average growth of 3.47%, peaked at 4.39% and slowed down to 2.39%.

IT industry has ceased to be a growth engine for the economy and its place can be taken by a revitalized health industry. For one, there is the enormous demand to extend life and ease the pain of debilitating and uncommon illnesses that are growing with longer life spans. On the supply side, a welter of new technologies for drugs, medical equipment, sensors and information technologies are revitalizing the industry. While technologies for extending life and improving its quality raise costs, some of them can also be used to drastically lower cost in an environment that encourages competition. Thus the twin goals of lower costs and better quality accessible to everyone are possible if the industry is restructured to be driven by market forces.

Meanwhile, healthcare industry’s existing paradigm of medicinal chemistry has run its course. Much of the therapies for an ageing population are effectively palliative which alleviate chronic illnesses, like diabetes, Alzheimer’s and heart disease, without providing a definitive cure. As a result, the costs of health care are ballooning. Chronic illnesses account for a major share of the costs–75% of the total of $2.2 trillion spent on health care in 2007[i]–with 45% of Americans suffering from at least one chronic illness. At this point, the industry is a money sink—really a gorge—unable to meet the needs of the mass market for health. Consumers don’t get care worth their money.

At this point, there is very little effort at prevention of diseases to keep costs low despite the proliferation of devices to forewarn patients and take preemptive action. The onset of stroke, for example, can be detected early with sensing devices and the more damaging consequences such as brain damage can be preempted.

The costs of hospital stays are high and their costs are lowered with remote treatment of patients at home. Hospital readmissions are common among chronically ill patients and happen when patients are not monitored when they return home. A New England Health Institute study found that hospital readmissions can be reduced by 60% with remote monitoring compared to standard care and 50% when programs that include disease management with an estimated savings of $6.4 billion a year.

The inefficiencies in the industry are easily noticeable. Paperwork abounds which is conspicuous by its absence in other industries. Only 10% of the payments to vendors in the health sector are by electronic means which could save an estimated $30 billion. Electronic records can drastically lower administrative costs and help to create databases that can be analyzed to improve the quality of care.

The productivity of the medical personnel is low as a result of excessive paperwork. Nurses, for example, expend only 40% of their time attending to patients while the rest of the time is spent on paperwork[ii]. Radiologists spend 70% of their time in the analysis of images while the rest is reserved for paperwork[iii]. Regulatory compliance compulsions bog down skilled staff in paperwork and alternatives are hard to find when the law is inflexible.

Patients have to take time off from their work, suffer long and wasteful wait times to receive care for even minor conditions in this age of multi-media communications. It gets worse for rural folks who have to trudge to urban hospitals. Care of dangerous criminals is hazardous when they have to be moved from prisons to hospitals. When emergencies are involved, patients have to be moved from one hospital to another when the required specialized personnel are not available. An estimated 2.2 million trips between emergency rooms happen for this reason. Entrepreneurs could very well help to bring care to patients with remote care technologies and video conferencing. The estimated savings from video assisted consultations replacing transportation from one emergency room to another are $537 million per annum.

Today, technology is available to substitute for doctor’s skills and hospitalization while maintaining the quality of care without forcing down doctor’s salaries by fiat. An example is Angioplasty which replaces expensive surgery for the treatment of coronary disease[iv]. The entire procedure can be completed by a skilled technician or a nurse at a much lower cost than a specialist in heart disease. Innovations of this nature can expand the health market by tapping into the latent mass market for health care.

Therapies are not customized for each patient and often many different options are tried in several hospitals to no avail. The problem is the inability of doctors to pinpoint the root cause of the disease. Normal radiological methods don’t necessarily help to diagnose a condition. Lately, DNA sequencing methods, such as those offered by Illumina, helped to diagnose the causes of an inflamed bowel of a six year old child after a hundred surgeries failed. The DNA sequencing data can also be mined to glean insights about the most effective treatments.

Health care costs are high also due to the structure of the industry. In the health delivery segment of the industry, hospital monopolies in local areas and regions raise the cost of services. Consolidations of hospitals are believed to add $12 billion a year to costs. The value chain could be broken up and individual tasks performed with much greater efficiency. Physician owned specialized hospitals are fierce competitors of General Hospitals especially in states with lighter regulation. In the thirty two states in which they do exist, physician hospitals are the top performers in nineteen of them and among the top in thirteen[v]. Regulation prevents them from operating in the remaining twenty of them.

In the risk management component of the industry, insurance plans are not portable outside a state or a medical group. Most insurance plans are paid by employers or by the Government for the elderly. Users of insurance plans don’t have much incentive to shop for health care or realize savings by participating in wellness programs. Health Savings Accounts (HSAs) put consumers of healthcare in charge of their choices by letting them spend out of their own pocket or from savings reserved for the purpose. They are expected to drive down costs by shopping for alternatives, evaluation of the value they are receiving for the money they pay and by becoming more aware of their own health risks. The evidence about the experience with HSAs is mixed; enrollment has increased, premiums are lower compared to traditional plans[vi] and their rate of increases is lower. The differences are smaller when adjustments are made for the age and the risk of the buyers of consumer plans and the traditional plans[vii]. On the other hand, the evidence on quality of care received by buyers of HSAs is mixed as is the care they take to search for their best option and lifestyle choices to lower their healthcare costs.

Innovation will begin to drive the health care industry and become the growth engine for the economy when costs and the corresponding quality of service are transparent and comparable. Currently, costs are estimated for individual departments and not by conditions and individual patients[viii]. It’s only then that consumers will shop for the best offering and switch to the most competitive offering and entrepreneurs will know where they can find opportunities to compete with incumbents. Even by rough measures, the quality of care, for the same expense, varies enormously—by a factor of 82[ix]

The search for the most desired doctors and therapies, with the best value to offer for the price paid, is arduous. Price for medical services varies enormously and comparisons are hard to make because of a maze of discounts offered. Most consumers of health have no incentive to shop for health care as their insurance plans are paid for by their employers. Aggregation of information and electronic searching will lower the costs of finding the best doctor and therapies. As patients search for the best option, it will also stimulate competition among vendors. The experience with high-deductible plans does show that health costs drop when users do shop around, compare before they buy healthcare.

The inefficiencies in the health industry are an enormous potential opportunity for growth which will receive another fillip when new bio-tech products move up their S-curve. For the USA, with its lead in medical innovation, there is also an opportunity to expand overseas where again health systems are largely inefficient. The key to tapping the latent opportunity is an environment that encourages technological innovation and competition. Employer paid health insurance plans or Government paid health insurance will be most effective when they encourage individuals to buy their own insurance plans and manage their own risk. For the rest, billions of consumers shopping for health care will drive down costs much faster than any group insurance or a Government department. At this point in time, the health marketplace offers very little data to make comparisons or even the flexibility to switch from one source to another. The incentive for cost reduction and to adopt new innovations will be the greatest when vendors deal with a more competitive environment.

A new paradigm in health care will be possible in such a competitive environment. Genomics, together with bio-technology, nanotechnology and medical devices, health IT, and sensors, will help to launch treatments that have for very long been elusive and affect large masses of people. Additionally, they are customized for each patient, who could be immune to standard treatments, and they have a much greater focus on prevention. These technologies are also able to shorten the duration of the illness, replace damaged organs and reverse the degenerative effects of ageing. Hypertension and high blood pressure, for example, affects 20% of Americans and the available cures generally require life-long treatments. The discovery of the relationship between genetic mutations and extremes in blood pressure[x] has improved an understanding of possible preventive treatments for high blood pressure. Diagnostic tools, which detect changes in the protein structure, will anticipate the onset of dreaded diseases like cancer and enable advanced action before the disease becomes incurable, Early results are accurate and clinical adoption is expected to come soon[xi].

[i] “Almanac of Chronic Disease”, 2009.

[ii] “Growth and Renewal in the USA: Retooling America’s economic engine”, McKinsey Global Institute, 2011

[iii] “Strategic Flexibility for the Health Plan Industry”, Deloitte.

[iv] “Will Disruptive Innovations Cure Health Care?” by Clayton M. Christensen, Richard Bohmer, and John Kenagy in Harvard Business Review.

[v] Consumer Reports quoted in “Why America needs more Physician hospitals”, The Senior Center for Health and Security, August 2009

[vi] “Generally, premiums for CDHPs were lower than premiums for non-CDHPs in all years except 2005, when premiums for HRA plans were higher than premiums for non-CDHPs. By 2009, annual premiums averaged $4,274 for HRA-based plans, $4,517 for HSA-eligible plans, and $4,902 for non-CDHP plans. Note that the $4,517 premium for HSA-eligible plans includes an average $688 employer contribution to the HSA account. Hence, premiums for HSA-eligible coverage were $3,829 for employee-only coverage in 2009”, in “What Do We Really Know About Consumer-Driven Health Plans?”, by Paul Fronstin, Employee Benefit Research Institute

[vii] op cit

[viii] “Discovering—and lowering—the real costs of health care”, by Michael Porter, Harvard Business Review,

[ix] “When and how provider competition can help improve health care delivery”, McKinsey.

[x] “The future of the biomedical industry in an era of globalization”, Kellogg School of Management, 2008

[xi] “In this case Proteomics is being used as a diagnostic tool and early data from the projects have been very positive with the computer software managing to identify 100% of ovarian cancer samples (when compared to a healthy sample) and 96% of prostate cancer samples. With these positive results it is surely only a matter of time before diagnostic Proteomics is seen in a clinical setting”, in “Could Proteomics be the future of cancer therapy?

.

Posted on

YouTube competes with commercial TV

By Kishore Jethanandani

Youtube’s harum-scarum expansion of goofy user-generated content is giving way to first steps towards professional content on premium TV channels. Sports content is the linchpin for commercial TV and will likely light the blaze of the trail to its long anticipated disruption by online social TV.

Youtube’s spending on premium channels has more than doubled in 2013 over 2012. The “grants” received by producers of content for premium channels increased from $100 million dollars in 2012 to $250 dollars in 2013 according to the numbers cited by a Needham Insights report. Content producers set aside all their revenues up to the amount of the grant to payback YouTube.

YouTube has not yet purchased broadcasting rights from mainstream sports clubs but it is able to gain the rights for broadcasting niche sports or a geography outside the main centers of the game. Skydiving is largely unknown outside small groups of hobbyists but found an audience of 8 million on Youtube when Felix Baumgartner did a sound barrier breaking jump. Professional Bull Riders Association (PBR), a game known to few people, saw an advantage in an all-digital strategy, in collaboration with YouTube, in order to expand its reach. Major League Baseball streams games on Youtube outside of the USA, NBA streams its minor league sports and the London Olympics expanded its reach to Asia.

The future of broadcasting rights for major league games is caught in a limbo as conflicting forces drive them in opposite directions. Cable companies need exclusive broadcasting rights to retain their customers and they are bidding at a higher rate to keep them. On the other hand, the hold of cable companies on broadcasting rights is tenuous as the audience shifts online to access content on mobile devices or any device. The number of households without any television jumped to 5 million in 2012 compared to 2 million in 2007.

Meanwhile, Youtube is able to cater to the demand by sports clubs to keep fans engaged with highlights of the game, interviews with sports men and women and peeks into the backrooms. One such Youtube channel is Love Football that has snippets of the game with captivating content like the footage on the goals scored, moments of suspense and moments of skillful maneuvering and reviews of the news.

Some of the more successful and competitive sports broadcasters supplement their live programming with partnerships with Youtube. ESPN, the leading sports broadcaster, has 1.2 million subscribers to its Youtube channel. Fox Sports, the upcoming challenger, has over 69,000 subscribers and sees their growth as a priority. They are looking to gain an edge with the interaction of fans with the content.

The demand for streaming media will only grow as 30 percent of cable subscribers have expressed their willingness to switch to streaming media. Youtube will be able to tip the balance of forces in the broadcasting industry in its favor when a high speed broadband network like the one Google is making available in Kansas is more widely available across the USA for superior quality programming.

Youtube’s premium channels will bring some of the benefits of commercial TV, such as ease of discovery, while keeping the fun of personalization of the Internet world. Not all of the freewheeling ways of the Internet world will be lost as fans will still be able to contribute their content in conversations, within the communities created by sports clubs, with the added benefit of convenient cataloging. A tough battle over broadcasting rights looms as the numbers of fans participating on Internet channels rises rapidly.

A version of this post was published in Digital Canvas Retail hosted by UBM Techweb

Posted on

Is Video Analytics an answer to Google Analytics for retail stores?

By Kishore Jethanandani

The competitive battle between brick-and-mortar stores and on-line stores seemed like a no contest. E-commerce stores have the clinching advantage in analytics. Video analytics may well help stores disprove that presumption.

E-commerce sites learn so much from the footprints of visitors on their sites with the analytics tools. Video analytics will potentially capture not only the quantitative data on shoppers but also the demographics such as ethnicity and even suggestive psychographic data, from the unguarded expressions of shoppers engaging with products and services, advertising and content, for better targeting.

Accepted notions of display, store planning and merchandising begin to change with the insights received from the analysis of video and related data. Tim Callan, the CMO of RetailNext, a leading video analytics company based in San Francisco, related the story of a retail store that wanted to position its signature products so that they would be conspicuous to shoppers. The conventional wisdom was that those products should be placed at the entry of the store. “Video analytics revealed that there are dead zones near the door entry where shoppers make way for people so they have enough space to enter the shop and miss the products placed there,” Mr. Tim Callan recounted.

Another case led to a substantial redesign of the store as a result of findings from video analytics. Like most lifestyle retail stores, it had a section for shoes. Typically, shoes are displayed on walls for customers to scan before they choose. Most stores place benches next to the wall where customers try out the shoes. “Video analytics revealed that the benches discouraged the shoppers from spending enough time looking at shoes on the wall,” Mr. Callan revealed. “The store decided to reserve separate spaces for the wall and the benches and increased the dwell time on shoes by five times,” Mr. Callan added.

Insights on customer behavior revealed by the qualitative data proved to be especially useful for Gordmans (Mr. Callan disputes the name of the store), a mid-western department store, as a guide to customer service decisions according to a report by the Economist. A camera embedded in Mannequins helped to spot a pattern—the tendency of Asian customers to visit at the same hours of the day. The store responded by placing Asian employees to serve them.

Merchandising decisions can be fine-tuned to increase conversions and improve outcomes from cross-selling. Retailers like Urban Outfitters draw on insights uncovered from correlations between the data on time spent eyeing products and sales realized at the counters. Customers, for example, could have entered the store to buy custom jewelry and may well be in the mood to buy perfumes. If they spot perfumes in an adjacent display, they are more likely to buy them. Today’s cameras take a 360 degree view of stores and will capture behavior of this nature to provide pointers to cross-selling opportunities.

The critical advantage of video analytics tools for stores is the ability to use analytical reports in real-time and feed advertisements, content and offers to customers while they are still in the store. Immersive Software’s Cara software, for example, processes data, in real-time, from face detection cameras to determine the gender and age of shoppers and change the advertisements that are more likely to appeal to the observed profile of visitors to the stores. “The data helps to personalize the experience for customers such as by placing customers in the advertisements they see,” said Jason Sosa, the CEO of Immersive Software.

Early adopter American Apparel recovered from a near death experience after it restructured its stores and deployed video analytics among other technologies. A store that faced the prospect of bankruptcy in mid-2012 is now profitable.

Stores have a chance to make shopping fun as they understand the shoppers’ behavior a lot better with video analytics. Shoppers could well end up loving it more than the yawn of executing commands on-line.

A version of this post was published in the now defunct Digital Canvas Retail hosted by UBM Techweb