Category Archives: Machine Learning

Wayne Blodwell, founder and chief exec of The Programmatic Advisory & The Programmatic University, battles through the buzzwords to explain why custom machine learning can help you unlock differentiation and regain a competitive edge.

Back in the day, simply having programmatic on plan was enough to give you a competitive advantage and no one asked any questions. But as programmatic has grown, and matured (84.5% of US digital display spend is due to be bought programmatically in 2020, the UK is on track for 92.5%), whats next to gain advantage in an increasingly competitive landscape?

Machine Learning

[noun]

The use and development of computer systems that are able to learn and adapt without following explicit instructions, by using algorithms and statistical models to analyse and draw inferences from patterns in data.

(Oxford Dictionary, 2020)

Youve probably head of machine learning as it exists in many Demand Side Platforms in the form of automated bidding. Automated bidding functionality does not require a manual CPM bid input nor any further bid adjustments instead, bids are automated and adjusted based on machine learning. Automated bids work from goal inputs, eg achieve a CPA of x or simply maximise conversions, and these inputs steer the machine learning to prioritise certain needs within the campaign. This tool is immensely helpful in taking the guesswork out of bids and the need for continual bid intervention.

These are what would be considered off-the-shelf algorithms, as all buyers within the DSP have access to the same tool. There is a heavy reliance on this automation for buying, with many even forgoing traditional optimisations for fear of disrupting the learnings and holding it back but how do we know this approach is truly maximising our results?

Well, we dont. What we do know is that this machine learning will be reasonably generic to suit the broad range of buyers that are activating in the platforms. And more often than not, the functionality is limited to a single success metric, provided with little context, which can isolate campaign KPIs away from their true overarching business objectives.

Custom machine learning

Instead of using out of the box solutions, possibly the same as your direct competitors, custom machine learning is the next logical step to unlock differentiation and regain an edge. Custom machine learning is simply machine learning that is tailored towards specific needs and events.

Off-the-self algorithms are owned by the DSPs; however, custom machine learning is owned by the buyer. The opportunity for application is growing, with leading DSPs opening their APIs and consoles to allow for custom logic to be built on top of existing infrastructure. Third party machine learning partners are also available, such as Scibids, MIQ & 59A, which will develop custom logic and add a layer onto the DSPs to act as a virtual trader, building out granular strategies and approaches.

With this ownership and customisation, buyers can factor in custom metrics such as viewability measurement and feed in their first party data to align their buying and success metrics with specific business goals.

This level of automation not only provides a competitive edge in terms of correctly valuing inventory and prioritisation, but the transparency of the process allows trust to rightfully be placed with automation.

Custom considerations

For custom machine learning to be effective, there are a handful of fundamental requirements which will help determine whether this approach is relevant for your campaigns. Its important to have conversations surrounding minimum event thresholds and campaign size with providers, to understand how much value you stand to gain from this path.

Furthermore, a custom approach will not fix a poor campaign. Custom machine learning is intended to take a well-structured and well-managed campaign and maximise its potential. Data needs to be inline for it to be adequately ingested and for real insight and benefit to be gained. Custom machine learning cannot simply be left to fend for itself; it may lighten the regular day to day load of a trader, but it needs to be maintained and closely monitored for maximum impact.

While custom machine learning brings numerous benefits to the table transparency, flexibility, goal alignment its not without upkeep and workflow disruption. Levels of operational commitment may differ depending on the vendors selected to facilitate this customisation and their functionality, but generally buyers must be willing to adapt to maximise the potential that custom machine learning holds.

Find out more on machine learning in a session The Programmatic University are hosting alongside Scibids on The Future Of Campaign Optimisation on 17 September. Sign up here.

See the original post here:
What is 'custom machine learning' and why is it important for programmatic optimisation? - The Drum

Artificial intelligence can be used to diagnose cancer, predict suicide, and assist in surgery. In all these cases, studies suggest AI outperforms human doctors in set tasks. But when something does go wrong, who is responsible?

Theres no easy answer, says Patrick Lin, director of Ethics and Emerging Sciences Group at California Polytechnic State University. At any point in the process of implementing AI in healthcare, from design to data and delivery, errors are possible. This is a big mess, says Lin. Its not clear who would be responsible because the details of why an error or accident happens matters. That event could happen anywhere along the value chain.

Design includes creation of both hardware and software, plus testing the product. Data encompasses the mass of problems that can occur when machine learning is trained on biased data, while deployment involves how the product is used in practice. AI applications in healthcare often involve robots working with humans, which further blurs the line of responsibility.

Responsibility can be divided according to where and how the AI system failed, says Wendall Wallace, lecturer at Yale Universitys Interdisciplinary Center for Bioethics and the author of several books on robot ethics. If the system fails to perform as designed or does something idiosyncratic, that probably goes back to the corporation that marketed the device, he says. If it hasnt failed, if its being misused in the hospital context, liability would fall on who authorized that usage.

Surgical Inc., the company behind the Da Vinci Surgical system, has settled thousands of lawsuits over the past decade. Da Vinci robots always work in conjunction with a human surgeon, but the company has faced allegations of clear error, including machines burning patients and broken parts of machines falling into patients.

Some cases, though, are less clear-cut. If diagnostic AI trained on data that over-represents white patients then misdiagnoses a Black patient, its unclear whether the culprit is the machine-learning company, those who collected the biased data, or the doctor who chose to listen to the recommendation. If an AI program is a black box, it will make predictions and decisions as humans do, but without being able to communicate its reasons for doing so, writes attorney Yavar Bathaee in a paper outlining why the legal principles that apply to humans dont necessarily work for AI. This also means that little can be inferred about the intent or conduct of the humans that created or deployed the AI, since even they may not be able to foresee what solutions the AI will reach or what decisions it will make.

The difficulty in pinning the blame on machines lies in the impenetrability of the AI decision-making process, according to a paper on tort liability and AI published in the AMA Journal of Ethics last year. For example, if the designers of AI cannot foresee how it will act after it is released in the world, how can they be held tortiously liable?, write the authors. And if the legal system absolves designers from liability because AI actions are unforeseeable, then injured patients may be left with fewer opportunities for redress.

AI, as with all technology, often works very differently in the lab than in a real-world setting. Earlier this year, researchers from Google Health found that a deep-learning system capable of identifying symptoms of diabetic retinopathy with 90% accuracy in the lab caused considerable delays and frustrations when deployed in real life.

Despite the complexities, clear responsibility is essential for artificial intelligence in healthcare, both because individual patients deserve accountability, and because lack of responsibility allows mistakes to flourish. If its unclear whos responsible, that creates a gap, it could be no one is responsible, says Lin. If thats the case, theres no incentive to fix the problem. One potential response, suggested by Georgetown legal scholar David Vladeck, is to hold everyone involved in the use and implementation of the AI system accountable.

AI and healthcare often work well together, with artificial intelligence augmenting the decisions made by human professionals. Even as AI develops, these systems arent expected to replace nurses or automate human doctors entirely. But as AI improves, it gets harder for humans to go against machines decisions. If a robot is right 99% of the time, then a doctor could face serious liability if they make a different choice. Its a lot easier for doctors to go along with what that robot says, says Lin.

Ultimately, this means humans are ceding some authority to robots. There are many instances where AI outperforms humans, and so doctors should defer to machine learning. But patient wariness of AI in healthcare is still justified when theres no clear accountability for mistakes. Medicine is still evolving. Its part art and part science, says Lin. You need both technology and humans to respond effectively.

See the original post here:
When AI in healthcare goes wrong, who is responsible? - Quartz

Strategic growth, latest insights, developmental trends in Global & Regional Machine Learning Courses Market with post-pandemic situations are reflected in this study. End to end Industry analysis from the definition, product specifications, demand till forecast prospects are presented. The complete industry developmental factors, historical performance from 2015-2027 is stated. The market size estimation, Machine Learning Courses maturity analysis, risk analysis, and competitive edge is offered. The segmental market view by types of products, applications, end-users, and top vendors is stated. Market drivers, restraints, opportunities in Machine Learning Courses industry with the innovative and strategic approach is offered. Machine Learning Courses product demand across regions like North America, Europe, Asia-Pacific, South and Central America, Middle East, and Africa is analyzed. The emerging segments, CAGR, revenue accumulation, feasibility check is specified.

Know more about this report or browse reports of your interest here:https://www.reportscheck.com/shop/global-machine-learning-courses-market-research-report-2015-2027-of-major-types-applications-and-competitive-vendors-in-top-regions-and-countries/#sample-request

COVID-19 has greatly impacted different Machine Learning Courses segments causing disruptions in the supply chain, timely product deliveries, production processes, and more. Post pandemic era the Machine Learning Courses industry will emerge with completely new norms, plans and policies, and development aspects. There will be new risk factors involved along with sustainable business plans, production processes, and more. All these factors are deeply analyzed by Reports Check's domain expert analysts for offering quality inputs and opinions.

Check out the complete table of contents, segmental view of this industry research report:https://www.reportscheck.com/shop/global-machine-learning-courses-market-research-report-2015-2027-of-major-types-applications-and-competitive-vendors-in-top-regions-and-countries/#table-of-contents

The qualitative and quantitative information is formulated in Machine Learning Courses report. Region-wise or country-wise reports are exclusively available on clients' demand with Reports Check. The market size estimation, Machine Learning Courses industry's competition, production capacity is evaluated. Also, import-export details, pricing analysis, upstream raw material suppliers, and downstream buyers analysis is conducted.

Receive complete insightful information with past, present and forecast situations of Global Machine Learning Courses Market and Post-Pandemic Status. Our expert analyst team is closely monitoring the industry prospects and revenue accumulation. The report will answer all your queries as well as you can make a custom request with free sample report.

A full-fledged, comprehensive research technique is used to derive Machine Learning Courses market's quantitative information. The gross margin, Machine Learning Courses sales ratio, revenue estimates, profits, and consumer analysis is provided. The complete global Machine Learning Courses market size, regional, country-level market size, & segmentation-wise market growth and sales analysis are provided. Value chain optimization, trade policies, regulations, opportunity analysis map, & marketplace expansion, and technological innovations are stated. The study sheds light on the sales growth of regional and country-level Machine Learning Courses market.

The company overview, total revenue, Machine Learning Courses financials, SWOT analysis, and product launch events are specified. We offer competitor analysis under the competitive landscape section for every competitor separately. The report scope section provides in-depth analysis of overall growth, leading companies with their successful Machine Learning Courses marketing strategies, market contribution, recent developments, and historic and present status.

Segment 1: Describes Machine Learning Courses market overview with definition, classification, product picture, Machine Learning Courses specifications

Segment 2: Machine Learning Courses opportunity map, market driving forces, restraints, and risk analysis

Segment 3:Competitive landscape view, sales, revenue, gross margin, pricing analysis, and global market share analysis

Segment 4:Machine Learning Courses Industry fragments by key types, applications, top regions, countries, top companies/manufacturers and end-users

Segment 5:Regional level growth, sales, revenue, gross margin from 2015-2020

Segment 6,7,8:Country-level sales, revenue, growth, market share from 2015-2020

Segment 9:Market sales, size, and share by each product type, application, and regional demand with production and Machine Learning Courses volume analysis

Segment 10:Machine Learning Courses Forecast prospects situations with estimates revenue generation, share, growth rate, sales, demand, import-export, and more

Segment 11 & 12:Machine Learning Courses sales and marketing channels, distributor analysis, customers, research findings, conclusion, and analysts views and opinions

Click to know more about our company and service offerings:https://www.reportscheck.com/shop/global-machine-learning-courses-market-research-report-2015-2027-of-major-types-applications-and-competitive-vendors-in-top-regions-and-countries/

An efficient research technique with verified and reliable data sources is what makes us stand out of the crowd. Excellent business approach, diverse clientele, in-depth competitor analysis, and efficient planning strategy is what makes us stand out of the crowd. We cater to different factors like technological innovations, economic developments, R&D, and mergers and acquisitions are specified. Credible business tactics and extensive research is the key to our business which helps our clients in profitable business plans.

Contact Us:

Olivia Martin

Email: [emailprotected]

Website:www.reportscheck.com

Phone: +1(831)6793317

See the original post here:
Global Machine Learning Courses Market Research Report 2015-2027 of Major Types, Applications and Competitive Vendors in Top Regions and Countries -...

One of the top 10 trends in data and analytics this year as leaders navigate the covid-19 world, according to Gartner, is augmented data management." Its the growing use of tools with ML/AI to clean and prepare robust data for AI-based analytics. Companies are currently striving to go digital and derive insights from their data, but the roadblock is bad data, which leads to faulty decisions. In other words: garbage in, garbage out.

I was talking to a university dean the other day. It had 20,000 students in its database, but only 9,000 students had actually passed out of the university," says Deleep Murali, co-founder and CEO of Bengaluru-based Zscore. This kind of faulty data has a cascading effect because all kinds of decisions, including financial allocations, are based on it.

Zscore started out with the idea of providing AI-based business intelligence to global enterprises. But the startup soon ran into a bigger problem: the domino effect of unreliable data feeding AI engines. We realized we were barking up the wrong tree," says Murali. Then we pivoted to focus on automating data checks."

For example, an insurance company allocates a budget to cover 5,000 hospitals in its database but it turns out that one-third of them are duplicates with a slight alteration in name. So far in pilots weve run for insurance companies, we showed $35 million in savings, with just partial data. So its a huge problem," says Murali.

EXPENSE & EFFORT

This is what prompted IBM chief Arvind Krishna to reveal that the top reason for its clients to halt or cancel AI projects was their data. He pointed out that 80% of an AI project involves collecting and cleansing data, but companies were reluctant to put in the effort and expense for it.

That was in the pre-covid era. Whats happening now is that a lot of companies are keen to accelerate their digital transformation. So customer traction is picking up from banks and insurance companies as well as the manufacturing sector," says Murali.

Data analytics tends to be on the fringes of a companys operations, rather than its core. Zscores product aims to change that by automating data flow and improving its quality. Use cases differ from industry to industry. For example, a huge drain on insurance companies is false claims, which can vary from absurdities like male pregnancies and braces for six-month-old toddlers to subtler cases like the same hospital receiving allocations under different names.

We work with a leading insurance company in Australia and claims leakage is its biggest source of loss. The moment you save anything in claims, it has a direct impact on revenue," says Murali. Male pregnancies and braces for six-month-olds seem like simple leaks but companies tend to ignore it. Legacy systems and rules havent accounted for all the possibilities. But now a claim comes to our system and multiple algorithms spot anything suspicious. Its a parallel system to the existing claims processing system."

For manufacturing companies, buggy inventory data means placing orders for things they dont need. For example, there can be 15 different serial numbers of spanners. So you might order a spanner thats well-stocked, whereas the ones really required dont show up. Companies lose 12-15% of their revenue each because of data issues such as duplicate or excessive inventory," says Murali.

These problems have got exacerbated in the age of AI where algorithms drive decision-making. Companies typically lack the expertise to prepare data in a way that is suitable for machine-learning models. How data is labelled and annotated plays a huge role. Hence, the need for supervised machine learning from tech companies like Zscore that can identify bad data and quarantine it.

TO THE ROOTS

Semantics and context analysis and studying manual processes help develop industry- or organization-specific solutions. So far 80-90% of data work has been manual. What we do is automate identification of data ingredients, data workflows and root cause analysis to understand whats wrong with the data," says Murali.

A couple of years ago, Zscore got into cloud data management multinational NetApps accelerator programme in Bengaluru. This gave it a foothold abroad with a NetApp client in Australia. It also opened the door to working with large financial institutions.

The Royal Commission of Australia, which is the equivalent of RBI, had come down hard on the top four banks and financial institutions for passing on faulty information. Its report said decisions had to be based on the right data and gave financial institutions 18 months to show progress. This became motivation for us because these were essentially data-oriented problems," says Murali.

Malavika Velayanikal is a consulting editor with Mint. She tweets @vmalu.

Subscribe to Mint Newsletters

* Enter a valid email

* Thank you for subscribing to our newsletter.

Follow this link:
The confounding problem of garbage-in, garbage-out in ML - Mint

According to a new report by PMMI Business Intelligence, artificial intelligence (AI) and machine learning is the area of automation technology with the greatest capacity for expansion. This technology can optimize individual processes and functions of the operation; manage production and maintenance schedules; and, expand and improve the functionality of existing technology such as vision inspection.

While AI is typically aimed at improving operation-wide efficiency, machine learning is directed more toward the actions of individual machines; learning during operation, identifying inefficiencies in areas such as rotation and movement, and then adjusting processes to correct for inefficiencies.

The advantages to be gained through the use of AI and machine learning are significant. One study released by Accenture and Frontier Economics found that by 2035, AI-empowered technology could increase labor productivity by up to 40%, creating an additional $3.8 trillion in direct value added (DVA) to the manufacturing sector.

See it Live at PACK EXPO Connects Nov. 9-13: End-of-Line Automation without Capital Expenditure, by Pearson Packaging Systems. Preview the Showroom Here.

However, only 1% of all manufacturers, both large and small, are currently utilizing some form of AI or machine learning in their operations. Most manufacturers interviewed said that they are trying to gain a better understanding of how to utilize this technology in their operations, and 45% of leading CPGs interviewed predict they will incorporate AI and/or machine learning within ten years.

A plant manager at a private label SME reiterates AI technology is still being explored, stating: We are only now talking about how to use AI and predict it will impact nearly half of our lines in the next 10 years.

While CPGs forecast that machine learning will gain momentum in the next decade, the near-future applications are likely to come in vision and inspection systems. Manufacturers can utilize both AI and machine learning in tandem, such as deploying sensors to key areas of the operation to gather continuous, real-time data on efficiency, which can then be analyzed by an AI program to identify potential tweaks and adjustments to improve the overall process.

See it Live at PACK EXPO Connects Nov. 9-13: Reduce costs and improve product quality in adhesive application of primary packaging, by Robatech USA Inc. Preview the Showroom Here.

And, the report states, that while these may appear to be expensive investments best left for the future, these technologies are increasingly affordable and offer solutions that can bring measurable efficiencies to smart manufacturing. In the days of COVID-19, gains to labor productivity and operational efficiency may be even more timely.

To access this FREE report and learn more about automation in operations, download below.

Source: PMMI Business Intelligence, Automation Timeline: The Drive Toward 4.0 Connectivity in Packaging and Processing

PACK EXPO Connects November 9-13. Now more than ever, packaging and processing professionals need solutions for a rapidly changing world, and the power of the PACK EXPO brand delivers the decision makers you need to reach. Attendeeregistrationis open now.

The rest is here:
Is Wide-Spread Use of AI & Machine Intelligence in Manufacturing Still Years Away? - Automation World

This article is part of ourreviews of AI research papers, a series of posts that explore the latest findings in artificial intelligence.

The Game of Life is a grid-based automaton that is very popular in discussions about science, computation, and artificial intelligence. It is an interesting idea that shows how very simple rules can yield very complicated results.

Despite its simplicity, however, the Game of Life remains a challenge to artificial neural networks, AI researchers at Swarthmore College and the Los Alamos National Laboratory have shown in a recent paper. Titled, Its Hard for Neural Networks To Learn the Game of Life, their research investigates how neural networks explore the Game of Life and why they often miss finding the right solution.

Their findings highlight some of the key issues with deep learning models and give some interesting hints at what could be the next direction of research for the AI community.

British mathematician John Conway invented the Game of Life in 1970. Basically, the Game of Life tracks the on or off statethe lifeof a series of cells on a grid across timesteps. At each timestep, the following simple rules define which cells come to life or stay alive, and which cells die or stay dead:

Based on these four simple rules, you can adjust the initial state of your grid to create interesting stable, oscillating, and gliding patterns.

For instance, this is whats called the glider gun.

You can also use the Game of Life to create very complex pattern, such as this one.

Interestingly, no matter how complex a grid becomes, you can predict the state of each cell in the next timestep with the same rules.

With neural networks being very good prediction machines, the researchers wanted to find out whether deep learning models could learn the underlying rules of the Game of Life.

There are a few reasons the Game of Life is an interesting experiment for neural networks. We already know a solution, Jacob Springer, a computer science student at Swarthmore College and co-author of the paper, told TechTalks. We can write down by hand a neural network that implements the Game of Life, and therefore we can compare the learned solutions to our hand-crafted one. This is not the case in.

It is also very easy to adjust the flexibility of the problem in the Game of Life by modifying the number of timesteps in the future the target deep learning model must predict.

Also, unlike domains such as computer vision or natural language processing, if a neural network has learned the rules of the Game of Life it will reach 100 percent accuracy. Theres no ambiguity. If the network fails even once, then it is has not correctly learned the rules, Springer says.

In their work, the researchers first created a small convolutional neural network and manually tuned its parameters to be able to predict the sequence of changes in the Game of Lifes grid cells. This proved that theres a minimal neural network that can represent the rule of the Game of Life.

Then, they tried to see if the same neural network could reach optimal settings when trained from scratch. They initialized the parameters to random values and trained the neural network on 1 million randomly generated examples of the Game of Life. The only way the neural network could reach 100 percent accuracy would be to converge on the hand-crafted parameter values. This would imply that the AI model had managed to parameterize the rules underlying the Game of Life.

But in most cases the trained neural network did not find the optimal solution, and the performance of the network decreased even further as the number of steps increased. The result of training the neural network was largely affected by the chosen set training examples as well as the initial parameters.

Unfortunately, you never know what the initial weights of the neural network should be. The most common practice is to pick random values from a normal distribution, therefore settling on the right initial weights becomes a game of luck. As for the training dataset, in many cases, it isnt clear which samples are the right ones, and in others, theres not much of a choice.

For many problems, you dont have a lot of choice in dataset; you get the data that you can collect, so if there is a problem with your dataset, you may have trouble training the neural network, Springer says.

In machine learning, one of the popular ways to improve the accuracy of a model that is underperforming is to increase its complexity. And this technique worked with the Game of Life. As the researchers added more layers and parameters to the neural network, the results improved and the training process eventually yielded a solution that reached near-perfect accuracy.

But a larger neural network also means an increase in the cost of training and running the deep learning model.

On the one hand, this shows the flexibility of large neural networks. Although a huge deep learning model might not be the most optimal architecture to address your problem, it has a greater chance of finding a good solution. But on the other, it proves that there is likely to be a smaller deep learning model that can provide the same or better resultsif you can find it.

These findings are in line with The Lottery Ticket Hypothesis, presented at the ICLR 2019 conference by AI researchers at MIT CSAIL. The hypothesis suggested that for each large neural network, there are smaller sub-networks that can converge on a solution if their parameters have been initialized on lucky, winning values, thus the lottery ticket nomenclature.

The lottery ticket hypothesis proposes that when training a convolutional neural network, small lucky subnetworks quickly converge on a solution, the authors of the Game of Life paper write. This suggests that rather than searching extensively through weight-space for an optimal solution, gradient-descent optimization may rely on lucky initializations of weights that happen to position a subnetwork close to a reasonable local minima to which the network converges.

While Conways Game of Life itself is a toy problem and has few direct applications, the results we report here have implications for similar tasks in which a neural network is trained to predict an outcome which requires the network to follow a set of local rules with multiple hidden steps, the AI researchers write in their paper.

These findings can apply to machine learning models used logic or math solvers, weather and fluid dynamics simulations, and logical deduction in language or image processing.

Given the difficulty that we have found for small neural networks to learn the Game of Life, which can be expressed with relatively simple symbolic rules, I would expect that most sophisticated symbol manipulation would be even more difficult for neural networks to learn, and would require even larger neural networks, Springer said. Our result does not necessarily suggest that neural networks cannot learn and execute symbolic rules to make decisions, however, it suggests that these types of systems may be very difficult to learn, especially as the complexity of the problem increases.

The researchers further believe that their findings apply to other fields of machine learning that do not necessarily rely on clear-cut logical rules, such as image and audio classification.

For the moment, we know that, in some cases, increasing the size and complexity of our neural networks can solve the problem of poorly performing deep learning models. But we should also consider the negative impact of using larger neural networks as the go-to method to overcome impasses in machine learning research. One outcome can be greater energy consumption and carbon emissions caused from the compute resources required to train large neural networks. On the other hand, it can result in the collection of larger training datasets instead of relying on finding ideal distribution strategies across smaller datasets, which might not be feasible in domains where data is subject to ethical considerations and privacy laws. And finally, the general trend toward endorsing overcomplete and very large deep learning models can consolidate AI power in large tech companies and make it harder for smaller players to enter the deep learning research space.

We hope that this paper will promote research into the limitations of neural networks so that we can better understand the flaws that necessitate overcomplete networks for learning. We hope that our result will drive development into better learning algorithms that do not face the drawbacks of gradient-based learning, the authors of the paper write.

I think the results certainly motivate research into improved search algorithms, or for methods to improve the efficiency of large networks, Springer said.

Read more here:
Why neural networks struggle with the Game of Life - TechTalks