Key Takeaways

Hands-On Genetic Algorithms with Python by Eyal Wirsansky is a new book which explores the world of genetic algorithms to solve search, optimization, and AI-related tasks, and improve machine learning models. InfoQ interviewed Eyal Wirsansky about how genetic algorithms work and what they can be used for.

In addition to our interview, InfoQ was able to obtain a sample chapter which can be downloaded here.

InfoQ: How do genetic algorithms work?

Eyal Wirsansky: Genetic algorithms are a family of search algorithms inspired by the principles of evolution in nature. They imitate the process of natural selection and reproduction, by starting with a set of random solutions, evaluating each one of them, then selecting the better ones to create the next generation of solutions. As generations go by, the solutions we have get better at solving the problem. This way, genetic algorithms can produce high-quality solutions for various problems involving search, optimization, and learning. At the same time, their analogy to natural evolution allows genetic algorithms to overcome some of the hurdles encountered by traditional search and optimization algorithms, especially for problems with a large number of parameters and complex mathematical representations.

InfoQ: What type of problems do genetic algorithms solve?

Wirsansky: Genetic algorithms can be used for solving almost any type of problem, but they particularly shine where traditional algorithms cannot be used, or fail to produce usable results within a practical amount of time. For example, problems with very complex or non-existing mathematical representation, problems where the number of variables involved is large, and problems with noisy or inconsistent input data. In addition, genetic algorithms are better equipped to handle deceptive problems, where traditional algorithms may get trapped in a suboptimal solution.

Genetic algorithms can even deal with cases where there is no way to evaluate an individual solution by itself, as long as there is a way to compare two solutions and determine which of them is better. An example can be a machine learning-based agent that drives a car in a simulated race. A genetic algorithm can optimize and tune the agent by having different versions of it compete against each other to determine which version is better.

InfoQ: What are the best use cases for genetic algorithms?

Wirsansky: The most common use case is where we need to assemble a solution using a combination of many different available parts; we want to select the best combination, but the number of possible combinations is too large to try them all. Genetic algorithms can usually find a good combination within a reasonable amount of time. Examples can be scheduling personnel, planning of delivery routes, designing bridge structures, and also constructing the best machine learning model from many available building blocks, or finding the best architecture for a deep learning model.

Another interesting use case is where the evaluation is based on peoples opinion or response. For example, you can use the genetic algorithm approach to determine the design parameters for a web sitesuch as color palette, font size, and location of components on the pagethat will achieve the best response from customers, such as conversion or retention. This idea can also be used for genetic art artificially created paintings or music that prove pleasant to the human eye (or ear).

Genetic algorithms can also be used for ongoing optimizationcases where the best solution may change over time. The algorithm can run continuously within the changing environment and respond dynamically to these changes by updating the best solution based on the current generation.

InfoQ: How can genetic algorithms select the best subset of features for supervised learning?

Wirsansky: In many cases, reducing the number of featuresused as inputs for a model in supervised learningcan increase the models accuracy, as some of the features may be irrelevant or redundant. This will also result in a simpler, better generalizing model. But we need to figure out which are the features that we want to keep. As this comes down to finding the best combination of features out of a potentially immense number of possible combinations, genetic algorithms provide a very practical approach. Each potential solution is represented by a list of booleans, one for each feature.

The value of the boolean (0 or 1) represents the absence or presence of the corresponding feature. These lists of boolean values are used as genetic material, that can be exchanged between solutions when we mate them, or even mutated by flipping values randomly. Using these mating and mutation operations, we create new generations out of preceding ones, while giving an advantage to solutions that yielded better performing models. After a while, we can have some good solutions, each representing a subset of the features. This is demonstrated in Chapter 7 of the book (our sample chapter) with the UCI Zoo dataset using python code, where the best performance was achieved by selecting six particular features out of the original sixteen.

InfoQ: What are the benefits that we can get from using genetic algorithms with machine learning for hyperparameter tuning?

Wirsansky: Every machine learning model utilizes a set of hyperparametersvalues that are set before the training takes place and affect the way the learning is done. The combined effect of hyperparameters on the performance of the model can be significant. Unfortunately, finding the best combination of the hyperparameter valuesalso known as hyperparameter tuningcan be as difficult as finding a needle in a haystack.

Two common approaches are grid search and random search, each with its own disadvantages. Genetic algorithms can be used in two ways to improve upon these methods. One way is by optimizing the grid search, so instead of trying out every combination on the grid, we can search only a subset of combinations but still get a good combination. The other way is to conduct a full search over the hyperparameter space, as genetic algorithms are capable of handling a large number of parameters as well as different parameter types continuous, discrete and categorical. These two approaches are demonstrated in Chapter 8 of the book with the UCI Wine dataset using python code.

InfoQ: How can genetic algorithms be used in Reinforcement Learning?

Wirsansky: Reinforcement Learning (RL) is a very exciting and promising branch of machine learning, with the potential to handle complex, everyday-life-like tasks. Unlike supervised learning, RL does not present an immediate 'right/wrong' feedback, but instead provides an environment where a longer-term, cumulative reward is sought after. This kind of setting can be viewed as an optimization problem, another area where genetic algorithms excel.

As a result, genetic algorithms can be utilized for reinforcement learning in several different ways. One example can be determining the weights and biases of a neural network that interacts with its environment by mapping input values to output values. Chapter 10 of the book includes two examples of applying genetic algorithms to RL tasks, using the OpenAI Gym environments mountain-car and cart-pole.

InfoQ: What is bio-inspired computing?

Wirsansky: Genetic algorithms are just one branch within a larger family of algorithms called Evolutionary Computation, all inspired by Darwinian evolution. One particularly interesting member of this family is Genetic Programming, that evolves computer programs aiming to solve a specific problem. More broadly, as evolutionary computation techniques are based on various biological systems or behaviors, they can be considered part of the algorithm family known as Bio-inspired Computing.

Among the many fascinating members of this family are Ant Colony Optimizationimitating the way certain species of ants locate food and mark the paths to it, giving advantage to closer and richer locations of food; Artificial Immune Systems, capable of identifying and learning new threats, as well as applying the acquired knowledge and respond faster the next time a similar threat is detected; and Particle Swarm Optimization, based on the behavior of flocks of birds or schools of fish, where individuals within the group work together towards a common goal without central supervision.

Another related, broad field of computation is Artificial Life, involving systems and processes imitating natural life in different ways, such as computer simulations and robotic systems. Chapter 12 of the book includes two relevant Python-written examples, one solving a problem using genetic programming, and the otherusing particle swarm optimization.

Eyal Wirsansky is a senior software engineer, a technology community leader, and an artificial intelligence researcher and consultant. Eyal started his software engineering career as a pioneer in the field of voice over IP, and he now has over 20 years' experience of creating a variety of high-performing enterprise solutions. While in graduate school, he focused his research on genetic algorithms and neural networks. One outcome of his research is a novel supervised machine learning algorithm that combines the two. Eyal leads the Jacksonville (FL) Java user group, hosts the Artificial Intelligence for Enterprise virtual user group, and writes the developer-oriented artificial intelligence blog, ai4java.

Excerpt from:
Q&A on the Book Hands-On Genetic Algorithms with Python - InfoQ.com

With this release, Parasoft introduced artificial intelligence (AI) and machine learning (ML) in its reporting and analytics dashboard, extending its capabilities tolearn from both historical interactions with the code base and prior static analysis findings to predict relevance and prioritize the new findings. As a result, teams can increase productivity by eliminating tedious and time-consuming tasks. Adding even more efficiency to the modern development workflow is the new Visual Studio Code extension for static analysisand the Coverage Advisor, which uses advanced static code analysis to boost unit test creation.

Parasoft Remains at the Forefront of Leading-Edge Technology With the Release of C/C++test 2020.1

The latest release introduces capabilities to improve all aspects of delivery in software quality including the following integrations:

"The growing complexity of software systems forces organizations to modernize their toolchains and workflows. They're switching to Git feature branch workflowsapplying Docker containers and CMake. We see heavy IDEs being replaced with lightweight editors like Visual Studio Code, which are a better fit for projects containing millions of lines of code. Modern workflows, however, need to support requirements traceability to facilitate risk assessment and functional safety certifications," said Miroslaw Zielinski, Product Manager for Parasoft. "Our latest release of Parasoft C/C++test with Visual Studio Code extension, Requirements View, streamlined Docker deployments and traceability enhancements fits perfectly into this trend."

Parasoft continues to provide leading support for automated enforcement of industry coding guidelines with expanded coverage for updated security standards (2019 CWE Top 25 and On the Cusp), AUTOSAR C++14, and the new MISRA C 2012 Amendment 2. Keeping pace with guideline requirements ensures that Parasoft's tools continue to meet the changing needs of the industry.

About Parasoft

Parasoft continuously delivers quality software with its market-proven, integrated suite of automated software testing tools. Parasoft supports software organizations as they develop and deploy applications for the embedded, enterprise, and IoT markets. Parasoft's technologies reduce the time, effort, and cost of delivering secure, reliable, and compliant software by integrating static and runtime analysis; unit, functional, and API testing; and service virtualization. With our developer testing tools, manager reporting/analytics, and executive dashboarding, Parasoft enables organizations to succeed in today's most strategic ecosystems and development initiativesreal-time, safety-critical, cybersecure, agile, continuous testing, and DevOps.

SOURCE Parasoft

http://www.parasoft.com

Read more here:
Parasoft Unleashes Artificial Intelligence and Machine Learning to Accelerate Time to Market for the Safety-Critical Industry - PRNewswire

Author: CapitalWatch Staff

Canaan Inc. (Nasdaq: CAN) became publicly traded in New York in late November. It raised $90 million in its IPO, which Canaan's founder, chairman, and chief executive officer,Nangeng Zhang modestly called "a good start." Since that time, the company has met significant milestones in its mission to disrupt the supercomputing industry.

Operating since 2013, Hangzhou-based Canaan delivers supercomputing solutions tailored to client needs. The company focuses on the research and development of artificial intelligence (AI) technology specifically, AI chips, AI algorithms, AI architectures, system on a chip (SoC) integration, and chip integration. Canaan is also known as a top manufacturer of mining hardware in China the global leader in digital currency mining.

Since IPO, Canaan has made strides in accomplishing new projects, despite the hard-hit cross-industry crisis Covid-19 has caused worldwide. In a recent announcement, Canaan said it has developed a SaaS product which its partners can use to operate a cloud mining platform. Cloud mining allows users to mine digital currency without having to buy and maintain mining hardware and spend on electricity a trend that has been gaining popularity.

A Chip of the Future

Earlier this year, Canaan participatedat the 2020 International Consumer Electronics Show in Las Vegas, the world's largest tech show that attracts innovators from across the globe. Canaan impressed, showcasing its Kendryte K210 the world's first-ever RISC-V-based edge AI chip. The chip was released in September 2018 and has been in mass-production ever since.

K210 is Canaan's first chip. The AI chip is designed to carry out machine learning. The primary functions of the K210 are machine vision and semantic, which includes the KPU for computing convolutional neural networks and an APU for processing microphone array inputs. KPU is a general-purpose neural network processor with built-in convolution, batch normalization, activation, and pooling operations. The next-generation chip can detect faces and objects in real-time. Despite the high computing power, K210 consumes only 0.3W while other typical devices consume 1W.

More Than Just Chipping Away at Sales

As of September 30, 2019, Canaan has shipped more than 53,000 AI chips and development kits to AI product developers since release.

Currently, the sales of K210 are growing exponentially, according to CEO Zhang .

The company has moved quickly to the commercialization of chips, and developed modules, products and back-end SaaS, offering customers a "full flow of AI solutions."

Based on the first generation of K210, Canaan has formed critical strategic partnerships.

For example, the company launched joint projects with a leading AI algorithm provider, a top agricultural science and technology enterprise, and a well-known global soft drink manufacturer to deliversmart solutionsfor variousindustrialmarkets.

The Booming Blockchain Industry

Currently, Canaan is working under the development strategy of "Blockchain + AI." The company has made several breakthroughs in the blockchain and AI industry, including algorithm development and optimization, standard unit design, low-voltage and high-efficiency operation, high-performance design system and heat dissipation, etc. The company has also accumulated extensive experience in ASIC chip manufacturing, laying the foundation for its future growth.

Canaan released first-generation products based on Samsung's 8nm and SMIC's 14nm technologies in Q4 last year. The former has been shipped in Q1 this year, while the latter will be shipped in Q2. In February, it launched the second generation of the product which is more efficient, more cost-effective and offers better performance.

Currently, TSMC's 5nm technology is under development. This technology will further improve the company's machines' computing power and ensure Canaan's leading position in the blockchain hardware space.

"We are the leader in the industry," says Zhang.

Canaan's Covid-19 Strategy

During the Covid-19 outbreak, Canaan improved the existing face recognition access control system. The new software can detect and identify people wearing masks. At the same time, the intelligent attendance system has been integrated to assist human resource management

Integrating mining machine learning and AI, the K210 chip has been used on Avalon mining machine, which can identify and monitor potential network viruses through intelligent algorithms. The company will explore more innovative integration in the future.

Second-Generation Gem

In terms of AI, the company will launch the second-generation AI chip K510 this year. The design of its architecture has been "greatly" optimized, and the computing power is several times more robust than the K210. Later this year, Canaan will use this tech in areas including smart energy consumption, smart industrial parks, smart driving, smart retail, and smart finance.

Canaan's Cash

In terms of operating costs and R&D, the company's last-year operating cost dropped 13.3% year-on-year. In 2018 and 2019, Canaan recorded R&D expenses of 189.7 million yuan and 169 million yuan, respectively347 million yuan were used to incentivize core R&D personnel.

In addition, the company currently has more than 500 million yuan in cash ($70.5 million), will continue to operate under the "blockchain + AI" strategy, with a continued focus on the commercialization of its AI technology.

A Fruitful Future

Canaan began as a manufacturer of Bitcoin mining machines, but it has become more than that. In the short term, the Bitcoin halving cycle is approaching (Estimated to occur on May 11, 2020 CW); this should promote the sales of company's mining machine, In the long term, now a global leader in ASIC technology, Canaan could be in a unique position to meet supercomputing demand.

"Blockchain is a good start, but we'll go beyond that," says Zhang. "When a seed grows up to be a big tree, it will bear fruit."

So far, it has done just that. Just how high that "tree" can get remains to be seen, but one thing is certain: The Kendryte K210 chip will be the driving force fueling the company's growth.

Read more:
Canaan's Kendryte K210 and the Future of Machine Learning - CapitalWatch

LONDON--(BUSINESS WIRE)--As a part of its new article series that analyzes COVID-19s impact across industries, Quantzig, a premier analytics services provider, today announced the completion of its recent article Why Online Food Delivery Companies are Betting Big on AI and Machine Learning

The article also offers comprehensive insights on:

Human activity has slowed down due to the pandemic, but its impact on business operations has not. We offer transformative analytics solutions that can help you explore new opportunities and ensure business stability to thrive in the post-crisis world. Request a FREE proposal to gauge COVID-19s impact on your business.

With machine learning, you dont need to babysit your project every step of the way. Since it means giving machines the ability to learn, it lets them make predictions and also improve the algorithms on their own, says a machine learning expert at Quantzig.

After several years of being confined to technology labs and the pages of sci-fi books, today artificial intelligence (AI) and big data have become the dominant focal point for businesses across industries. Barely a day passes by without new magazine and paper articles, blog entries, and tweets about such advancements in the field of AI and machine learning. Having said that, its not very surprising that AI and machine learning in the food and beverage industry have played a crucial role in the rapid developments that have taken place over the past few years.

Talk to us to learn how our advanced analytics capabilities combined with proprietary algorithms can support your business initiatives and help you thrive in todays competitive environment.

Benefits of AI and Machine Learning

Want comprehensive solution insights from an expert who decodes data? Youre just a click away! Request a FREE demo to discover how our seasoned analytics experts can help you.

As cognitive technologies transform the way people use online services to order food, it becomes imperative for online food delivery companies to comprehend customer needs, identify the dents, and bridge gaps by offering what has been missing in the online food delivery business. The combination of big data, AI, and machine learning is driving real innovation in the food and beverage industry. Such technologies have been proven to deliver fact-based results to online food delivery companies that possess the data and the required analytics expertise.

At Quantzig, we analyze the current business scenario using real-time dashboards to help global enterprises operate more efficiently. Our ability to help performance-driven organizations realize their strategic and operational goals within a short span using data-driven insights has helped us gain a leading edge in the analytics industry. To help businesses ensure business continuity amid the crisis, weve curated a portfolio of advanced COVID-19 impact analytics solutions that not just focus on improving profitability but help enhance stakeholder value, boost customer satisfaction, and help achieve financial objectives.

Request more information to know more about our analytics capabilities and solution offerings.

About Quantzig

Quantzig is a global analytics and advisory firm with offices in the US, UK, Canada, China, and India. For more than 15 years, we have assisted our clients across the globe with end-to-end data modeling capabilities to leverage analytics for prudent decision making. Today, our firm consists of 120+ clients, including 45 Fortune 500 companies. For more information on our engagement policies and pricing plans, visit: https://www.quantzig.com/request-for-proposal

Here is the original post:
Quantzig Launches New Article Series on COVID-19's Impact - 'Understanding Why Online Food Delivery Companies Are Betting Big on AI and Machine...

Across industries, companies are applying artificial intelligence to their businesses, with mixed results. What separates the AI projects that succeed from the ones that dont often has to do with the business strategies organizations follow when applying AI, writes Wharton professor of operations, information and decisions Kartik Hosanagar in this opinion piece. Hosanagar is faculty director of Wharton AI for Business, a new Analytics at Wharton initiative that will support students through research, curriculum, and experiential learning to investigate AI applications. He also designed and instructs Wharton Onlines Artificial Intelligence for Business course.

While many people perceive artificial intelligence to be the technology of the future, AI is already here. Many companies across a range of industries have been applying AI to improve their businesses from Spotify using machine learning for music recommendations to smart home devices like Google Home and Amazon Alexa. That said, there have also been some early failures, such as Microsofts social-learning chatbot, Tay, which turned anti-social after interacting with hostile Twitter followers, and IBM Watsons inability to deliver results in personalized health care. What separates the AI projects that succeed from the ones that dont often has to do with the business strategies organizations follow when applying AI. The following strategies can help business leaders not only effectively apply AI in their organizations, but succeed in adapting it to innovate, compete and excel.

1. View AI as a tool, not a goal.

One pitfall companies might encounter in the process of starting new AI initiatives is that the concentrated focus and excitement around AI might lead to AI being viewed as a goal in and of itself. But executives should be cautious about developing a strategy specifically for AI, and instead focus on the role AI can play in supporting the broader strategy of the company. A recent report from MIT Sloan Management Review and Boston Consulting Group calls this backward from strategy, not forward from AI.

As such, instead of exhaustively looking for all the areas AI could fit in, a better approach would be for companies to analyze existing goals and challenges with a close eye for the problems that AI is uniquely equipped to solve. For example, machine learning algorithms bring distinct strengths in terms of their predictive power given high-quality training data. Companies can start by looking for existing challenges that could benefit from these strengths, as those areas are likely to be ones where applying AI is not only possible, but could actually disproportionately benefit the business.

The application of machine learning algorithms for credit card fraud detection is one example of where AIs particular strengths make it a very valuable tool in assisting with a longstanding problem. In the past, fraudulent transactions were generally only identified after the fact. However, AI allows banks to detect and block fraud in real time. Because banks already had large volumes of data on past fraudulent transactions and their characteristics, the raw material from which to train machine learning algorithms is readily available. Moreover, predicting whether particular transactions are fraudulent and blocking them in real time is precisely the type of repetitive task that an algorithm can do at a speed and scale that humans cannot match.

2. Take a portfolio approach.

Over the long term, viewing AI as a tool and finding AI applications that are particularly well matched with business strategy will be most valuable. However, I wouldnt recommend that companies pool all their AI resources into a single, large, moonshot project when they are first getting started. Rather, I advocate taking a portfolio approach to AI projects that includes both quick wins and long-term projects. This approach will allow companies to gain experience with AI and build consensus internally, which can then support the success of larger, more strategic and transformative projects later down the line.

Specifically, quick wins are smaller projects that involve optimizing internal employee touch points. For example, companies might think about specific pain points that employees experience in their day-to-day work, and then brainstorm ways AI technologies could make some of these tasks faster or easier. Voice-based tools for scheduling or managing internal meetings or voice interfaces for search are some examples of applications for internal use. While these projects are unlikely to transform the business, they do serve the important purpose of exposing employees, some of whom may initially be skeptics, to the benefits of AI. These projects also provide companies with a low-risk opportunity to build skills in working with large volumes of data, which will be needed when tackling larger AI projects.

The second part of the portfolio approach, long-term projects, is what will be most impactful and where it is important to find areas that support the existing business strategy. Rather than looking for simple ways to optimize the employee experience, long-term projects should involve rethinking entire end-to-end processes and potentially even coming up with new visions for what otherwise standard customer experiences could look like. For example, a long-term project for a car insurance company could involve creating a fully automated claims process in which customers can photograph the damage of their car and use an app to settle their claims. Building systems like this that improve efficiency and create seamless new customer experiences requires technical skills and consensus on AI, which earlier quick wins will help to build.

The skills needed for embarking on AI projects are unlikely to exist in sufficient numbers in most companies, making reskilling particularly important.

3. Reskill and invest in your talent.

In addition to developing skills through quick wins, companies should take a structured approach to growing their talent base, with a focus on both reskilling internal employees in addition to hiring external experts. Focusing on growing the talent base is particularly important given that most engineers in a company would have been trained in computer science before the recent interest in machine learning. As such, the skills needed for embarking on AI projects are unlikely to exist in sufficient numbers in most companies, making reskilling particularly important.

In its early days of working with AI, Google launched an internal training program where employees were invited to spend six months working in a machine learning team with a mentor. At the end of this time, Google distributed these experts into product teams across the company in order to ensure that the entire organization could benefit from AI-related reskilling. There are many new online courses to reskill employees in AI economically.

The MIT Sloan Management Review-BCG report mentioned above also found that, in addition to developing talent in producing AI technologies, an equally important area is that of consuming AI technologies. Managers, in particular, need to have skills to consult AI tools and act on recommendations or insights from these tools. This is because AI systems are unlikely to automate entire processes from the get-go. Rather, AI is likely to be used in situations where humans remain in the loop. Managers will need basic statistical knowledge in order to understand the limitations and capabilities of modern machine learning and to decide when to lean on machine learning models.

4. Focus on the long term.

Given that AI is a new field, it is largely inevitable that companies will experience early failures. Early failures should not discourage companies from continuing to invest in AI. Rather, companies should be aware of, and resist, the tendency to retreat after an early failure.

Historically, many companies have stumbled in their early initiatives with new technologies, such as when working with the internet and with cloud and mobile computing. The companies that retreated, that stopped or scaled back their efforts after initial failures, tended to be in a worse position long term than those that persisted. I anticipate that a similar trend will occur with AI technologies. That is, many companies will fail in their early AI efforts, but AI itself is here to stay. The companies that persist and learn to use AI well will get ahead, while those that avoid AI after their early failures will end up lagging behind.

AI shouldnt be abandoned given that the alternative, human decision-makers, are biased too.

5. Address AI-specific risks and biases aggressively.

Companies should be aware of new risks that AI can pose and proactively manage these risks from the outset. Initiating AI projects without an awareness of these unique risks can lead to unintended negative impacts on society, as well as leave the organizations themselves susceptible to additional reputational, legal, and regulatory risks (as mentioned in my book, A Humans Guide to Machine Intelligence: How Algorithms Are Shaping Our Lives and How We Can Stay in Control).

There have been many recent cases where AI technologies have discriminated against historically disadvantaged groups. For example, mortgage algorithms have been shown to have a racial bias, and an algorithm created by Amazon to assist with hiring was shown to have a gender bias, though this was actually caught by Amazon itself prior to the algorithm being used. This type of bias in algorithms is thought to occur because, like humans, algorithms are products of both nature and nurture. While nature is the logic of the algorithm itself, nurture is the data that algorithms are trained on. These datasets are usually compilations of human behaviors oftentimes specific choices or judgments that human decision-makers have previously made on the topic in question, such as which employees to hire or which loan applications to approve. The datasets are therefore made up of biased decisions from humans themselves that the algorithms learn from and incorporate. As such, it is important to note that algorithms are generally not creating wholly new biases, but rather learning from the historical biases of humans and exacerbating them by applying them on a much larger, and therefore even more damaging, scale.

AI shouldnt be abandoned given that the alternative, human decision-makers, are biased too. Rather, companies should be aware of the kinds of social harms that can result from AI technologies and rigorously audit their algorithms to catch biases before they negatively impact society. Proceeding with AI initiatives without an awareness of these social risks can lead to reputational, legal, and regulatory risks for firms, and most importantly can have extremely damaging impacts on society.

Originally posted here:
Five Strategies for Putting AI at the Center of Digital Transformation - Knowledge@Wharton