Learning Archives - Business Intelligence Info

Tag Archives: Learning

IBM releases Qiskit modules that use quantum computers to improve machine learning

April 11, 2021 Big Data

Join GamesBeat Summit 2021 this April 28-29. Register for a free or VIP pass today.

IBM is releasing Qiskit Machine Learning, a set of new application modules that’s part of its open source quantum software. The new feature is the latest expansion of the company’s broader effort to get more developers to begin experimenting with quantum computers.

According to a blog post by the Qiskit Applications Team, the machine learning modules promise to help optimize machine learning by using quantum computers for some parts of the process.

“Quantum computation offers another potential avenue to increase the power of machine learning models, and the corresponding literature is growing at an incredible pace,” the team wrote. “Quantum machine learning (QML) proposes new types of models that leverage quantum computers’ unique capabilities to, for example, work in exponentially higher-dimensional feature spaces to improve the accuracy of models.”

Rather than replacing current computer architectures, IBM is betting that quantum computers will gain traction in the coming years by taking on very specific tasks that are offloaded from a classic computing system to a quantum platform. AI and machine learning are among the areas where IBM has said it’s hopeful that quantum can make an impact.

To make quantum more accessible, last year IBM introduced an open source quantum programming framework called Qiskit. The company has said it has the potential to speed up some applications by 100 times.

In the case of machine learning, the hope is that a system that offloads tasks to a quantum system could accelerate the training time. However, challenges remain, such as how to get large data sets in and out of the quantum machine without adding time that would cancel out any gains by the quantum calculations.

Developers who use Qiskit to improve their algorithms will have access to test them on IBM’s cloud-based quantum computing platform.

VentureBeat

VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative technology and transact.

Our site delivers essential information on data technologies and strategies to guide you as you lead your organizations. We invite you to become a member of our community, to access:

up-to-date information on the subjects of interest to you
our newsletters
gated thought-leader content and discounted access to our prized events, such as Transform 2021: Learn More
networking features, and more

Become a member

Let’s block ads! (Why?)

Big Data – VentureBeat

AI Weekly: Continual learning offers a path toward more humanlike AI

April 10, 2021 Big Data

Join GamesBeat Summit 2021 this April 28-29. Register for a free or VIP pass today.

State-of-the-art AI systems are remarkably capable, but they suffer from a key limitation: statisticity. Algorithms are trained once on a dataset and rarely again, making them incapable of learning new information without retraining. This is as opposed to the human brain, which learns constantly, using knowledge gained over time and building on it as it encounters new information. While there’s been progress toward bridging the gap, solving the problem of “continual learning” remains a grand challenge in AI.

This challenge motivated a team of AI and neuroscience researchers to found ContinualAI, a nonprofit organization and open community of continual and lifelong learning enthusiasts. ContinualAI recently announced Avalanche, a library of tools compiled over the course of a year from over 40 contributors to make continual learning research easier and more reproducible. The group also hosts conference-style presentations, sponsors workshops and AI competitions, and maintains a repository of tutorial, code, and guides.

As Vincenzo Lomonaco, cofounding president and assistant professor at the University of Pisa, explains, ContinualAI is one of the largest organizations on a topic its members consider fundamental for the future of AI. “Even before the COVID-19 pandemic began, ContinualAI was funded with the idea of pushing the boundaries of science through distributed, open collaboration,” he told VentureBeat via email. “We provide a comprehensive platform to produce, discuss and share original research in AI. And we do this completely for free, for anyone.”

Even highly sophisticated deep learning algorithms can experience catastrophic learning or catastrophic interference, a phenomenon where deep networks fail to recall what they’ve learned from a training dataset. The result is that the networks have to be constantly reminded of the knowledge they’ve gained or risk becoming “stuck” with their most recent “memories.”

OpenAI research scientist Jeff Clune, who helped to cofound Uber AI Labs in 2017, has called catastrophic forgetting the “Achilles’ heel” of machine learning and believes that solving it is the fastest path to artificial general intelligence (AGI). Last February, Clune coauthored a paper detailing ANML, an algorithm that managed to learn 600 sequential tasks with minimal catastrophic forgetting by “meta-learning” solutions to problems instead of manually engineering solutions. Separately, Alphabet’s DeepMind has published research suggesting that catastrophic forgetting isn’t an insurmountable challenge for neural networks. And Facebook is advancing a number of techniques and benchmarks for continual learning, including a model that it claims is effective in preventing the forgetting of task-specific skills.

But while the past several years have seen a resurgence of research into the issue, catastrophic forgetting largely remains unsolved, according to Keiland Cooper, a cofounding member of ContinualAI and a neuroscience research associate at the University of California, Irvine. “The potential of continual learning exceeds catastrophic forgetting and begins to touch on more interesting questions of implementing other cognitive learning properties in AI,” Cooper told VentureBeat. “Transfer learning is one example, where when humans or animals learn something previously, sometimes this learning can be applied to a new context or aid learning in other domains … Even more alluring is that continual learning is an attempt to push AI from narrow, savant-like systems to broader, more general ones.”

Even if continual learning doesn’t yield the sort of AGI depicted science fiction, Cooper notes that there are immediate advantages to it across a range of domains. Cutting-edge models are being trained on increasingly larger datasets in search of better performance, but this training comes at a cost — whether waiting weeks for training to finish or the impact of the electricity usage on the environment.

“Say you run a certain AI organization that built a natural language model that was trained over weeks on 45 terabytes of data for a few million dollars,” Cooper explained. “If you want to teach that model something new, well, you’d very likely have to start from scratch or risk overwriting what it had already learned, unless you added continual learning additions to the model. Moreover, at some point, the cost to store that data will be exceedingly high for an organization, or even impossible. Beyond this, there are many cases where you can only see the data once and so retraining isn’t even an option.”

While the blueprint for a continual learning AI system remains elusive, ContinualAI aims to connect researchers and stakeholders interested in the area and support and provide a platform for projects and research. It’s grown to over 1,000 members in the three years since its founding.

“For me personally, while there has been a renewed interest in continual learning in AI research, the neuroscience of how humans and animals can accomplish these feats is still largely unknown,” Cooper said. “I’d love to see more of an interaction with AI researchers, cognitive scientists, and neuroscientists to communicate and build upon each of their fields ides towards a common goal of understanding one of the most vital aspects of learning and intelligence. I think an organization like ContinualAI is best positioned to do just that, which allows for the sharing of ideas without the boundaries of the academic or industry walls, siloed fields, or distant geolocation.”

Beyond the mission of dissemination information about continual learning, Lomonaco believes that ContinualAI has the potential to become a reference points for a more inclusive and collaborative way of doing research in AI. “Elite university and private company labs still work mostly behind close doors, [but] we truly believe in inclusion and diversity rather than selective elitiarity. We favor transparency and open-source rather than protective IP licenses. We make sure anyone has access to the learning resources she needs to achieve her potential.”

For AI coverage, send news tips to Kyle Wiggers — and be sure to subscribe to the AI Weekly newsletter and bookmark our AI channel, The Machine.

Thanks for reading,

Kyle Wiggers

AI Staff Writer

VentureBeat

VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative technology and transact.

Our site delivers essential information on data technologies and strategies to guide you as you lead your organizations. We invite you to become a member of our community, to access:

up-to-date information on the subjects of interest to you
our newsletters
gated thought-leader content and discounted access to our prized events, such as Transform 2021: Learn More
networking features, and more

Become a member

Let’s block ads! (Why?)

Big Data – VentureBeat

Announcing Plant-Based Machine Learning

April 5, 2021 BI News and Info

Around this time last year, we announced that we’d expanded our AI and ML capabilities to pets, truly making machine learning paw-ssible for anyone, regardless of whether you’re human or not.

And while we were proud of that effort, some of us wondered if we could go further. There are, after all, other groups of living creatures on this planet. Although we initially considered an expansion into the fungal kingdom, early experiments only resulted in a lot of moldy circuit boards.

But we didn’t let that stop us! We kept working at it, and this year, we’re excited to continue expanding the reach of AI and ML by providing members of the plant kingdom with the ability to run their own machine learning models, using the cloud-based infrastructure of RapidMiner Go. We were so pleased with how much we’ve expanded machine learning beyond humans that we went so far as to revise our understanding of our mission statement:

Our new toolset is based on a proprietary, state-of-the-art plant/computer interface called Lexical-Epicotyl Access for Plants and Herbs (LEAPH) that our green friends can use to access the Internet to find data and then use RapidMiner Go to build and train models.

Let’s take a look at a few of the plants that we worked with during prototyping to give you a sense of the kinds of models that plants might be interested in building and how they can use this technology to have a positive impact on our world.

Potato (Solanum tuberosum)

Potatoes get a bad rap—carbophobes hate them, they’re not infrequently forgotten in a bottom drawer where they turn into brown slush, and they’re even the butt of internet jokes, comparing them to crummy cameras and computers.

But with LEAPH, they can fight back against their haters! Using RapidMiner Go, our resident potato plants were able to build a model that could detect negative sentiment about potatoes on various websites and forums. Their next step is to build a model that can respond to try and discourage people from saying things like photos have “potato quality”.

So if you see anyone complaining about language that’s disparaging to potatoes online, you’ve probably run into the first potato-built sentiment analysis and influencer bot.

Avocado (Persea americana)

Avocados are superstars of the produce aisle—high in healthy fats, great on toast, and popular at parties in the form of guacamole.

But avocados have their problems, too. No avocado wants to be purchased only to be taken home and tossed in the trash when its all-too-short eating window comes and goes without anyone taking notice. This is especially concerning to the avocados since their unpredictable ripening times drive up costs for end consumers, potentially impacting their popularity and thus the number of avocados grown each year.

With our LEAPH interface, a grove of avocado trees was able to model the development of their fruits—including subjective measures like “how brown under the skin are they” and “do they have any of those weird thread things” that’s inaccessible to us humans on the outside—and try to create a model that lets people know when the avocados are ready to be eaten.

Unfortunately, initial testing didn’t result in a model that was able to accurately predict when an avocado was ripe with better than chance accuracy. However, the avocado trees are hopeful that they’ll eventually crack their own code, providing more predictability about their ripeness, and in turn, driving down costs so that restaurants can stop charging extra for guac.

Kale (Brassica oleracea)

Kale has exploded in popularity in the last decade or so, and we have no idea why—and neither does kale! With LEAPH and RapidMiner Go, kale was able to plumb the Internet to look for evidence and chart its rise to health-food hero. Although celebrity endorsements seem to have played a role, the main driver is that people love to talk about how health they are on social media. Since eating kale is seen as a key part of a healthy lifestyle, more social media has driven more kale sales.

And the kale ain’t complainin’!

Wrapping Up

These are just a few examples of the kinds of impacts that RapidMiner can have when it puts the power of AI and ML in anyone’s hands—even plants. We’re excited to see what other plants come up with as we roll out our LEAPH interfaces in the coming weeks.

Our new plant-based options are sure to rock the machine learning world to its roots, as is our new slogan: RapidMiner Go: literally so easy a potato can do it.

Let’s block ads! (Why?)

RapidMiner

Major flaws found in machine learning for COVID-19 diagnosis

March 24, 2021 Big Data

Join Transform 2021 for the most important themes in enterprise AI & Data. Learn more.

A coalition of AI researchers and health care professionals in fields like infectious disease, radiology, and ontology have found several common but serious shortcomings with machine learning made for COVID-19 diagnosis or prognosis.

After the start of the global pandemic, startups like DarwinAI, major companies like Nvidia, and groups like the American College of Radiology launched initiatives to detect COVID-19 from CT scans, X-rays, or other forms of medical imaging. The promise of such technology is that it could help health care professionals distinguish between pneumonia and COVID-19 or provide more options for patient diagnosis. Some models have even been developed to predict if a person will die or need a ventilator based on a CT scan. However, researchers say major changes are needed before this form of machine learning can be used in a clinical setting.

Researchers assessed more than 2,200 papers and, through a process of removing duplicates and irrelevant titles, narrowed results down to 320 papers that underwent a full text review for quality. Finally, 62 papers were deemed fit to be part of what authors refer to as a systematic review of published research and preprints shared on open research paper repositories like arXiv, bioRxiv, and medRxiv.

Of those 62 papers included in the analysis, roughly half made no attempt to perform external validation of training data, did not assess model sensitivity or robustness, and did not report the demographics of people represented in training data.

“Frankenstein” datasets, the kind made with duplicate images obtained from other datasets, were also found to be a common problem, and only one in five COVID-19 diagnosis or prognosis models shared their code so others can reproduce results claimed in literature.

“In their current reported form, none of the machine learning models included in this review are likely candidates for clinical translation for the diagnosis/prognosis of COVID-19,” the paper reads. “Despite the huge efforts of researchers to develop machine learning models for COVID-19 diagnosis and prognosis, we found methodological flaws and many biases throughout the literature, leading to highly optimistic reported performance.”

The research was published last week as part of the March issue of Nature Machine Intelligence by researchers from the University of Cambridge and University of Manchester. Other common issues they found with machine learning models developed using medical imaging data was virtually no assessment for bias and generally being trained without enough images. Nearly every paper reviewed was found to be at high or uncertain risk of bias; only six were considered at low risk of bias.

Publicly available datasets also commonly suffered from lower quality image formats and weren’t large enough to train reliable AI models. Researchers used the checklist for artificial intelligence in medical imaging (CLAIM) and radiomics quality score (RQS) to help assess the datasets and models.

“The urgency of the pandemic led to many studies using datasets that contain obvious biases or are not representative of the target population, for example, pediatric patients. Before evaluating a model, it is crucial that authors report the demographic statistics for their datasets, including age and sex distributions,” the paper reads. “Higher-quality datasets, manuscripts with sufficient documentation to be reproducible and external validation are required to increase the likelihood of models being taken forward and integrated into future clinical trials to establish independent technical and clinical validation as well as cost-effectiveness.”

Other recommendations suggested by the group of AI researchers and health care professionals include ensuring reproducibility of model performance results spelled out in research papers and considering how datasets are assembled and put together.

In other news at the intersection of COVID-19 and machine learning, earlier this week the Food and Drug Administration (FDA) approved emergency use authorization of a machine learning-based screening device which the agency says is the first approved in the U.S.

VentureBeat

VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative technology and transact.

Our site delivers essential information on data technologies and strategies to guide you as you lead your organizations. We invite you to become a member of our community, to access:

up-to-date information on the subjects of interest to you
our newsletters
gated thought-leader content and discounted access to our prized events, such as Transform 2021: Learn More
networking features, and more

Become a member

Let’s block ads! (Why?)

Big Data – VentureBeat

Not Getting the Most From Your Model Ops? Why Businesses Struggle With Data Science and Machine Learning

March 18, 2021 TIBCO Spotfire

TIBCO ModelOps scaled e1615222627944 696x366 Not Getting the Most From Your Model Ops? Why Businesses Struggle With Data Science and Machine Learning

Reading Time: 2 minutes

Companies have begun to recognize the value of integrating data science (DS) and machine learning (ML) across their organization to reap the benefits of the advanced analytics they can provide. As such, DS/ML has seen a surge in popularity and usage as businesses have invested heavily in this technology.

However, there’s a distinct difference between investing in DS/ML and managing to successfully gain tangible business value from that investment, and that’s where organizations are running into problems.

The Results Are in: Businesses Struggle With DS/ML Deployment Across the Board

We recently performed a global survey across 18 countries and 22 industries, including over a hundred business leaders and executives, more than half of which were in the C-Suite.

Of those respondents, just 14 percent reported that they are currently operationalizing DS/ML. Within that 14 percent, 24 percent can only use it in one functional area, far below the potential innovative capability of the technology.

Why are so few organizations able to follow through with model ops adoption? What are the barriers keeping businesses from operationalizing data science and machine learning?

The Devil’s in the Data

According to the survey results, while a lack of talented data scientists to build the models was listed in the top ten obstacles to DS/ML adoption, it was only cited by about 16 percent of respondents. On the other hand, seven of these ten, including the top four, were all data-related. Issues with data security, data privacy, data prep, and data access, in particular, were all cited by between 27 to 38 percent of respondents.

While there are many other issues to contend with, including lack of management and financial support and a clear integration strategy, security compliance and data privacy concerns are clearly a significant barrier when it comes to operationalizing DS/ML.

Why Overcoming These Problems are Critical for Innovation

Data scientists can develop as many models as they want for a business, but if they don’t get deployed, then they aren’t providing any value. For the modern digital business to have any hope of keeping up with the competition, model ops is a vital tool that can allow them to effectively operationalize DS/ML models, putting them into production and applying them to streaming, real-time data, edge applications, and more.

For a more in-depth breakdown of our survey results, you can check out our full ebook now. And if you’re ready to move past insights and into action, you can download our four-step guide to finding out what it takes to operationalize data science within your organization and get a leg-up on the competition.

Let’s block ads! (Why?)

The TIBCO Blog

OctoML raises $28M for machine learning deployment optimization

March 17, 2021 Big Data

Join Transform 2021 for the most important themes in enterprise AI & Data. Learn more.

It’s well known that most businesses face challenges deploying AI in production, and that’s led to the rise of markets serve those needs. In the latest news of advance for such a company, OctoML today raised a $ 28 million Series B funding round.

OctoML helps businesses accelerate and deploy AI and relies on the open-source technology Apache TVM machine learning compiler framework. The funding will be used for OctoML to continue building out products like the Octomizer platform and invest in the company’s go-to-market strategy and customer service teams.

“We started the TVM work as a research project at the University of Washington about five years ago, and all the key people in the project are part of, they all got their PhDs and are part of the company now,” OctoML CEO and cofounder Luis Ceze told VentureBeat. “We’re focused on making inference fast on any hardware, and support cloud and edge deployments.”

Last month, OctoML joined more than 20 startups who have banded together to create the AI Infrastructure Alliance, an effort involving startups like Algorithmia and Determined AI for interoperability between the offerings from AI startups and advance alternatives to popular cloud AI services.

The $ 28 million funding was led by Addition Capital led the round with participation from existing investors Madrona Venture Group and Amplify Partners.

OctoML has raised $ 47 million to date. A $ 3.9 million seed funding round was held in October 2019.

VentureBeat

VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative technology and transact.

Our site delivers essential information on data technologies and strategies to guide you as you lead your organizations. We invite you to become a member of our community, to access:

up-to-date information on the subjects of interest to you
our newsletters
gated thought-leader content and discounted access to our prized events, such as Transform 2021: Learn More
networking features, and more

Become a member

Let’s block ads! (Why?)

Big Data – VentureBeat

Why Choose RapidMiner for Your Data Science & Machine Learning Software?

March 6, 2021 BI News and Info

Recent recognition received from Gartner and RapidMiner’s end-users

If you follow RapidMiner’s blog, you know that we shy away from self-promotion. We’re much more interested in lending expertise that can help your data science efforts—model creation best practices, cutting-edge industry examples of AI/ML usage, and recommendations on how to overcome common business challenges that can prevent operationalization of machine learning.

However, we’ll make exceptions when our product team gets some much-deserved recognition. Their work empowers people to leverage machine learning more effectively, regardless of their role or technical skill-level. They make it easier to solve business problems and tackle new challenges through a more informed, analytical lens.

So, with that said…

RapidMiner Named a Visionary in this year’s Magic Quadrant for Data Science & Machine Learning Platforms

We’re excited to share that Gartner has named RapidMiner Visionary in this year’s Magic Quadrant for Data Science & Machine Learning Platforms! If you’re unfamiliar with Gartner Magic Quadrants, they offer an in-depth look at markets, the trends that define them, and key participants. Put simply, Magic Quadrant help companies get a sense for vendors’ strengths and weaknesses when they’re evaluating a purchase. By providing actionable insights and advice, they ensure that readers are making the most informed buying decisions they can.

We believe this year’s Magic Quadrant builds on the positive trends that we saw in 2020’s Gartner Peer Insights for Machine Learning and Data Science, which is the firm’s ranking of platforms in the market based on real end-user reviews. The feedback from RapidMiner users was so positive that we were awarded a Customers’ Choice Award, signifying that customers rated our platform strongly relative to competitors in the market. We’ve seen similar feedback from other reviewers on Forrester & G2 Crowd.

Over the past year, we’ve been focused on expanding our platform to meet the needs of larger enterprises. An examination of the challenges that these enterprises face when trying to use AI led us to invest in key areas like collaboration, governance, and explainable AI. The recent recognition we’ve received from both Gartner’s analysts & RapidMiner’s end-users validates our investment in those areas.

What Does it Mean to be Visionary?

The reason that we view this as such a key milestone isn’t just because we were recognized, but it’s because of the advancements that we’ve made in the realms of multi-persona collaboration, explainable AI, and model governance.

Multi-persona collaboration

In terms of collaboration, it’s always been a goal of ours to allow data scientists and business experts to work together to solve business problems because, while data scientists bring a wealth of knowledge to the table when it comes to building models and extracting insights, they don’t have experience working in functional areas every day, so they may not have as much context about business problems. By contrast, a Head of Production is in tune with the fact that they need to lower product defects & improve yield; they just don’t have the coding background to build models than can give them insight on where to start.

By allowing these two groups to work together in a single platform, companies using RapidMiner can ensure that they’re addressing the right problems and building technically sound models that will have strong business impact.

Explainable AI

In addition to getting teams to work better together, we’ve also been focused on ways to help you visualize and communicate the results of your data science work to others. We’re well aware of the fact that if can’t easily explain what your model is doing and how, it’s unlikely that you’re going to be able to get it implemented to have real business impact. And what better way to present model insights than with visualization?

By enabling users to visualize and explain the models that they’ve built, whether they’re in development or production, RapidMiner creates greater transparency, gives users full control over insights, and helps ensure that models can make it across the finish line and return dividends on the investment in machine learning.

Model governance

Lastly, we’ve invested a lot into helping companies establish secure, governable data science practices. This is rooted in the belief that no enterprise AI initiative is worth investing in if you can’t guarantee that your data is safe and properly governed. That’s why we’ve implemented auditable project-tracking, Single Sign-On (SSO), and strong identity and access management (IAM) capabilities within our platform, allowing admins to secure their AI pipeline, all in one place.

Looking Ahead

In addition to the three main points above, the Magic Quadrant also notes RapidMiner’s clear vision for what features need to be implemented in the future, as well as our ability to get those features right. For example, in 2020, we notably added capabilities that “enable users to perform automated feature engineering, and share and store features across an organization, thus enhancing reusability and reproducibility.”

Again, while the posts on this blog aren’t typically about RapidMiner, we’re proud to be recognized for our commitment to reinvent enterprise AI so that anyone has the power to positively shape the future. Check out everything Gartner has to say by reading the full report.

Gartner does not endorse any vendor, product or service depicted in its research publications, and does not advise technology users to select only those vendors with the highest ratings or other designation. Gartner research publications consist of the opinions of Gartner’s research organization and should not be construed as statements of fact. Gartner disclaims all warranties, express or implied, with respect to this research, including any warranties of merchantability or fitness for a particular purpose.

Gartner Magic Quadrant for Data Science and Machine Learning Platforms, Peter Krensky, Carlie Idoine, Erick Brethenoux, Pieter den Hamer, Farhan Choudhary, Afraz Jaffri, Shubhangi Vashisth,1^st March 2021.

Let’s block ads! (Why?)

RapidMiner

Facebook’s new computer vision model achieves state-of-the-art performance by learning from random images

March 4, 2021 Big Data

Join Transform 2021 for the most important themes in enterprise AI & Data. Learn more.

Facebook today announced an AI model trained on a billion images that ostensibly achieves state-of-the-art results on a range of computer vision benchmarks. Unlike most computer vision models, which learn from labeled datasets, Facebook’s generates labels from data by exposing the relationships between the data’s parts — a step believed to be critical to someday achieving human-level intelligence.

The future of AI lies in crafting systems that can make inferences from whatever information they’re given without relying on annotated datasets. Provided text, images, or another type of data, an AI system would ideally be able to recognize objects in a photo, interpret text, or perform any of the countless other tasks asked of it.

Facebook claims to have made a step toward this with a computer vision model called SEER, which stands for SElf-supERvised. SEER contains a billion parameters and can learn from any random group of images on the internet without the need for curation or annotation. Parameters, a fundamental part of machine learning systems, are the part of the model derived from historical training data.

New techniques

Self-supervision for vision is a challenging task. With text, semantic concepts can be broken up into discrete words, but with images, a model must decide for itself which pixel belongs to which concept. Making matters more challenging, the same concept will often vary between images. Grasping the variation around a single concept, then, requires looking at a lot of different images.

Facebook researchers found that scaling AI systems to work with complex image data required at least two core components. The first was an algorithm that could learn from a vast number of random images without any metadata or annotations, while the second was a convolutional network — ConvNet — large enough to capture and learn every visual concept from this data. Convolutional networks, which were first proposed in the 1980s, are inspired by biological processes, in that the connectivity pattern between components of the model resembles the visual cortex.

In developing SEER, Facebook took advantage of an algorithm called SwAV, which was borne out of the company’s investigations into self-supervised learning. SwAV uses a technique called clustering to rapidly group images from similar visual concepts and leverage their similarities, improving over the previous state-of-the-art in self-supervised learning while requiring up to 6 times less training time.

Facebook’s new computer vision model achieves state of the art performance by learning from random images

Above: A simplified schematic showing SEER’s model architecture.

Image Credit: Facebook

Training models at SEER’s size also required an architecture that was efficient in terms of runtime and memory without compromising on accuracy, according to Facebook. The researchers behind SEER opted to use RegNets, or a type of ConvNet model capable of scaling to billions or potentially trillions of parameters while fitting within runtime and memory constraints.

Facebook software engineer Priya Goyal said SEER was trained on 512 NVIDIA V100 GPUs with 32GB of RAM for 30 days.

The last piece that made SEER possible was a general-purpose library called VISSL, short for VIsion library for state-of-the-art Self Supervised Learning. VISSL, which Facebook is open-sourcing today, allows for self-supervised training with a variety of modern machine learning methods. The library facilitates self-supervised learning at scale by integrating algorithms that reduce the per-GPU memory requirement and increase the training speed of any given model.

Performance and future work

After pretraining on a billion public Instagram images, SEER outperformed the most advanced state-of-the-art self-supervised systems, Facebook says. SEER also outperformed models on tasks including object detection, segmentation, and image classification. When trained with just 10% of the examples in the popular ImageNet dataset, SEER still managed to hit 77.9% accuracy. And when trained with just 1%, SEER was 60.5% accurate.

When asked whether the Instagram users whose images were used to train SEER were notified or given an opportunity to opt out of the research, Goyal noted that Facebook informs Instagram account holders in its data policy that it uses information like pictures to support research, including the kind underpinning SEER. That said, Facebook doesn’t plan to share the images or the SEER model itself, in part because the model might contain unintended biases.

“Self-supervised learning has long been a focus for Facebook AI because it enables machines to learn directly from the vast amount of information available in the world, rather than just from training data created specifically for AI research,” Facebook wrote in a blog post. “Self-supervised learning has incredible ramifications for the future of computer vision, just as it does in other research fields. Eliminating the need for human annotations and metadata enables the computer vision community to work with larger and more diverse datasets, learn from random public images, and potentially mitigate some of the biases that come into play with data curation. Self-supervised learning can also help specialize models in domains where we have limited images or metadata, like medical imaging. And with no labor required up front for labeling, models can be created and deployed quicker, enabling faster and more accurate responses to rapidly evolving situations.”

VentureBeat

VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative technology and transact.

Our site delivers essential information on data technologies and strategies to guide you as you lead your organizations. We invite you to become a member of our community, to access:

up-to-date information on the subjects of interest to you
our newsletters
gated thought-leader content and discounted access to our prized events, such as Transform
networking features, and more

Become a member

Let’s block ads! (Why?)

Big Data – VentureBeat

Why machine learning strategies fail

February 26, 2021 Big Data

Most companies are struggling to develop working artificial intelligence strategies, according to a new survey by cloud services provider Rackspace Technology. The survey, which includes 1,870 organizations in a variety of industries, including manufacturing, finance, retail, government, and healthcare, shows that only 20 percent of companies have mature AI/machine learning initiatives. The rest are still trying to figure out how to make it work.

There’s no questioning the promises of machine learning in nearly every sector. Lower costs, improved precision, better customer experience, and new features are some of the benefits of applying machine learning models to real-world applications. But machine learning is not a magic wand. And as many organizations and companies are learning, before you can apply the power of machine learning to your business and operations, you must overcome several barriers.

Three key challenges companies face when integrating AI technologies into their operations are in the areas of skills, data, and strategy, and Rackspace’s survey paints a clear picture of why most machine learning strategies fail.

Machine learning is about data

Machine learning models live on compute resources and data. Thanks to a variety of cloud computing platforms, access to the hardware needed to train and run AI models has become much more accessible and affordable.

But data continues to remain a major hurdle in different stages of planning and adopting an AI strategy. Thirty-four percent of the respondents in the Rackspace survey stated poor data quality as the main reason for the failure of machine learning research and development, and another 31 percent said they lacked production-ready data.

This highlights one of the main hurdles when applying machine learning techniques to real-world problems. While the AI research community has access to many public datasets for training and testing their latest machine learning technologies, when it comes to applying those technologies to real applications, getting access to quality data is not easy. This is especially true in industrial, health, and government sectors, where data is often scarce or subject to strict regulations.

Data problems crop up again when machine learning initiatives move from the research to the production phase. Data quality remains the top barrier when it comes to using machine learning to extract valuable insights. Data engineering problems also pose a significant problem, such as data being siloed, lack of talent to connect disparate data sources, and not being fast enough to process data in a meaningful way.

Above: Data accounts for most key problems in gaining actionable insights from machine learning models (Source: Rackspace Technology)

Both startups and established companies suffer from data problems, though scale seems to be the key differentiator between the two, according to Jeff DeVerter, CTO of Rackspace Technology. “Startups tend to be constrained with not all the right resources to implement a quality data pipeline and consistently managing it over time,” DeVerter said to TechTalks in written comments. “Enterprises usually have scale on their side and with that comes the rigor that’s required.”

The best way companies can prepare for the data challenges of AI strategies is to do a full evaluation of their data infrastructure. Eliminating silos should be a key priority in every machine learning initiative. Companies should also have the right procedures for cleaning their data to improve the accuracy and performance of their machine learning models.

AI talent is still in high demand

The second area of struggle for most companies is access to machine learning and data science talent. According to Rackspace’s survey, lack of in-house expertise was the second biggest driver of failure in machine learning R&D initiatives. Lack of skill and difficulty in hiring was also a key barrier in adopting AI technologies.

Above: Many companies struggle with acquiring the talent to implement their AI strategies (Source: Rackspace Technology)

With machine learning and deep learning having reached mainstream use in production environments only recently, many smaller companies don’t have data scientists and machine learning engineers who can develop AI models.

And the average salary of data scientists and machine learning engineers matches those of experienced software engineers, which makes it difficult for many companies to put together a talented team that can lead its AI initiative.

While the shortage of machine learning and data science talent is well known, one thing that has gone mostly unnoticed is the need for more data engineers, the people who set up, maintain, and update databases, data warehouses, and data lakes. Per Rackspace’s figures, many initiatives fail because companies don’t have the talent to adapt their data infrastructure for machine learning purposes. Breaking down silos, migrating to cloud, setting up Hadoop clusters, and creating hybrid systems that can leverage the power of different platforms are some areas where companies are sorely lacking. And these shortcomings prevent them from making company-wide deployments of machine learning initiatives.

With the development of new machine learning and data science tools, the talent problem has become less intense. Google, Microsoft, and Amazon have launched platforms that make it easier to develop machine learning models. An example is Microsoft’s Azure Machine Learning service, which provides a visual interface with drag-and-drop components and makes it easier to create ML models without coding. Another example is Google’s AutoML, which automates the tedious process of hyperparameter tuning. While these tools are not a replacement for machine learning talent, they lower the barrier for people who want to enter the field and will enable many companies to reskill their tech talent for these growing fields.

“Lack of in-house data science talent is not the barrier it once was now that more of these services are able to use their own ML to help in this regard as well consulting firms having these talents on-staff,” DeVerter said.

Other developments in the field are the evolution of cloud storage and analysis platforms, which have considerably reduced the complexity of creating the seamless data infrastructures needed to create and run AI systems. An example is Google’s BigQuery, a cloud-based data warehouse that can run queries across vast amounts of data stored in various sources with minimal effort.

We’re also seeing growing compatibility and integration capabilities in machine learning tools, which will make it much easier for organizations to integrate ML tools into their existing software and data ecosystem.

Before entering an AI initiative, every organization must make a full evaluation of in-house talent, available tools, and integration possibilities. Knowing how much you can rely on your own engineers and how much it will cost you to hire talent will be a defining factor in the success or failure of your machine learning initiatives. Also, consider whether re-skilling is a possible course of action. If you can upskill your engineers to take on data science and machine learning projects, you will be better off in the long run.

Outsourcing AI talent

Another trend that has seen growth in recent years is the outsourcing of AI initiatives. Only 38 percent of the Rackspace survey respondents relied on in-house talent to develop AI applications. The rest were either fully outsourcing their AI projects or employing a combination of in-house and outsourced talent.

Above: Most companies rely on outside talent to plan and implement their AI initiatives (Source: Rackspace Technology)

There are now several companies that specialize in developing and implementing AI strategies. An example is C3.ai, an AI solutions provider that specializes in several industries. C3.ai provides AI tools on top of existing cloud providers such as Amazon, Microsoft, and Google. The company also provides AI consultancy and expertise to take customers step by step through the strategizing and implementation phases.

According to the Rackspace report: “A mature provider can bring everything from strategy to implementation to maintenance and support over time. Strategy can sidestep the areas where AI and machine learning efforts may lose momentum or get lost in complexity. Hands-on experts can also spare organizations from the messy work of cleanup and maintenance. Such expertise, taken together, can make all the difference in finally achieving success.”

It is worth noting, however, that fully turning over an organization’s AI strategy to outside providers can be a double-edged sword. A successful strategy requires close cooperation between AI specialists and subject matter experts from the company that is implementing the strategy.

“This is very similar to companies who move to a DevOps development methodology and attempt to outsource the entirety of the development. DevOps requires a close partnership between the developers, business analysts, and others in the business,” DeVerter said. “In the same way, AI projects require strategy and technical expertise — but also require a tight partnership with the business as well as leadership.”

Outsourcing AI talent must be done meticulously. While it can speed up the process of developing and implementing an AI strategy, you must make sure that your experts are fully involved in the process. Ideally, you should be able to develop your own in-house team of data scientists and machine learning engineers as you work with outside experts.

How do you evaluate your AI strategy?

Finally, another area that is causing much pain for companies embarking on an AI journey is forecasting the outcome and value of AI strategies. Given the application of machine learning being new to many areas, it’s hard to know in advance how long an AI strategy will take to plan and implement and what the return on investment will be. This in turn makes it difficult for innovators in organizations to get others on board when it comes to garnering support for AI initiatives.

Of the respondents of the Rackspace survey, 18 percent believed that a lack of clear business case was the main barrier to adopting AI strategies. Lack of commitment from executives was also among the top barriers. Lack of use cases and commitment from senior management show up again among the top challenges in the machine learning journey.

“AI often wanders around as a solution looking for a problem within organizations. I believe this is one of the greatest impediments to its wide-scale adoption within organizations,” DeVerter said. “As AI practitioners can demonstrate practical examples of how AI can benefit their specific company — leadership will further fund those activities. Like any business venture — leadership needs to know how it will either help them save or make money.”

Evaluating the outcome of AI initiatives is very difficult. According to the survey, the top-two key performance indicators (KPI) for measuring the success of AI initiatives were profit margins and revenue growth. Understandably, this focus on quick profits is partly due to the high costs of AI initiatives. According to the Rackspace survey, organizations spend a yearly average of $ 1.06 million on AI initiatives.

But while a good AI initiative should result in revenue growth and lower costs, in many cases, the long-term value of machine learning is the development of new use cases and products.

“Short-term financial gains can be myopic if they aren’t paired with a long-term strategy that can be funded by those short-term gains,” DeVerter said.

If you’re in charge of the AI initiative in your organization, make sure to clearly lay out the use cases, the costs, and the benefits of your AI strategy. Decision-makers should have a clear picture of what their company will be embarking on. They should understand the short-term benefits of investing in AI, but they should also know what they will gain in the long run.

Ben Dickson is a software engineer and the founder of TechTalks. He writes about technology, business, and politics. This post was originally published here as a series exploring the business of artificial intelligence.

VentureBeat

VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative technology and transact.

Our site delivers essential information on data technologies and strategies to guide you as you lead your organizations. We invite you to become a member of our community, to access:

up-to-date information on the subjects of interest to you
our newsletters
gated thought-leader content and discounted access to our prized events, such as Transform
networking features, and more

Become a member

Let’s block ads! (Why?)

Big Data – VentureBeat

How to know if federated learning should be part of your data strategy

February 13, 2021 Big Data

AI researchers and practitioners are developing and releasing new systems and methods at an extremely fast pace, and it can be difficult for enterprises to gauge which particular AI technologies are most likely to help their businesses. This article — the first part of a two-part series — will try to help you determine if federated learning (FL), a fairly new piece of privacy-preserving AI technology, is appropriate for a use case you have in mind.

FL’s core purpose is to enable use of AI in situations where data privacy or confidentiality concerns currently block adoption. Let’s unpack this a bit. The purpose of AI systems/methods/algorithms is to take data and autonomously create pieces of software, AI models, that transform data into actionable insight. Modern AI methods typically require a lot of data, collect all the necessary data at some central location, and run a learning algorithm on the data to learn/create the model.

However, when that data is confidential, siloed, and owned by different entities, gathering the data at a central location is not possible. Federated learning very cleverly gets around this problem by moving the learning algorithm or code to the data, rather than bringing the data to the code.

As an example, consider the problem of creating an AI model to predict whether patients have COVID-19 given their lung CT-scans. The data for this problem, CT-scans, are obviously confidential, and are owned by various different entities — hospitals and medical and research facilities. They are also stored all across the world under very various different jurisdictions. You want to combine all of that data because you want your AI models to be able to detect all the forms in which the disease has manifested in CT-scans. However, combining all this data in the one location is not possible for obvious data confidentiality and jurisdictional reasons.

Here’s how federated learning bypasses the confidentiality issue: A worker node, which is a computing system capable of machine learning, is deployed at each hospital or facility. This worker node has full access to the confidential data at the location. During federated learning, each such worker node creates a local AI model using the data at the facility and sends over the model to the central FL server. Note that the confidential data never leaves the client site — only the model does. A central server combines all of the insights from the local models into a single global model. This global model is sent back to all the local worker nodes, which now have insights from all the other worker nodes without having seen their data. When you iterate on these steps many times, you end up with a model that is, in many cases, equivalent to a model that would have been built if you’d trained it on all the data in the same place. (See a schematic illustration of the process below.)

Federated learning was initially developed by Google as a way to train the Android keyboard app (GBoard) to predict what the user will type next. Here the confidential data being used is the text that the user is typing. However, it turns out that the data confidentiality issue appears in many guises across industries. Indeed, you will face this problem if you are

* a maker of autonomous vehicle and you want to combine confidential video and image data from across your vehicles to build a better vision system,

* a bank and want to build a machine learning based anti-money laundering system by combining data from various jurisdictions, possibly even other banks,

* a supply chain platform provider and you want to build a better risk assessment or route optimization system by combining confidential sales data from multiple businesses,

* a cell service provider and you want to build a machine learning model to optimize routes by combining confidential data from cell towers,

* a consortium of farmers and you want to build a model to detect crop disease using confidential disease data from the members of your consortium,

* a consortium of additive manufacturers and you want to build AI-based process controllers and quality assurance systems using confidential build data from members of your consortium.

An important point to note about the above examples is that the owner of the data can be different units of the same organization but in different jurisdictions (as in the bank and cell service provider examples) or clients of the same organization (as in the autonomous vehicle maker and supply chain service provider examples), or completely independent units (as in the consortium of farmers and additive manufacturer examples above, and the hospitals/COVID-19 example described earlier).

You can use the following checklist to see if federated learning makes sense for you:

I will end by emphasizing that, if you are able to combine your data into a central location, you should probably go for that option (barring cases where you want to, for instance, future-proof your solution). Centralizing your data may make for better final model performance than you’ll be able to achieve with federated learning, and the effort required to deploy an AI solution will be significantly lower. However, if it turns out that FL is just what you need to remove your barrier to AI adoption, part two of this series will provide you with an overview of how to do that.

M M Hassan Mahmud is a Senior AI and Machine Learning Technologist at Digital Catapult, with a background in machine learning within academia and industry.

VentureBeat is always looking for insightful guest posts related to enterprise data technology and strategy.

VentureBeat

VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative technology and transact.

Our site delivers essential information on data technologies and strategies to guide you as you lead your organizations. We invite you to become a member of our community, to access:

up-to-date information on the subjects of interest to you
our newsletters
gated thought-leader content and discounted access to our prized events, such as Transform
networking features, and more

Become a member

Let’s block ads! (Why?)

Big Data – VentureBeat

Tag Archives: Learning

IBM releases Qiskit modules that use quantum computers to improve machine learning

VentureBeat

AI Weekly: Continual learning offers a path toward more humanlike AI

Announcing Plant-Based Machine Learning

Potato (Solanum tuberosum)

Avocado (Persea americana)

Kale (Brassica oleracea)

Wrapping Up

Major flaws found in machine learning for COVID-19 diagnosis

VentureBeat

Not Getting the Most From Your Model Ops? Why Businesses Struggle With Data Science and Machine Learning

The Results Are in: Businesses Struggle With DS/ML Deployment Across the Board

The Devil’s in the Data

Why Overcoming These Problems are Critical for Innovation

OctoML raises $28M for machine learning deployment optimization

VentureBeat

Why Choose RapidMiner for Your Data Science & Machine Learning Software?

Recent recognition received from Gartner and RapidMiner’s end-users

RapidMiner Named a Visionary in this year’s Magic Quadrant for Data Science & Machine Learning Platforms

What Does it Mean to be Visionary?

Multi-persona collaboration

Explainable AI

Model governance

Looking Ahead

Facebook’s new computer vision model achieves state-of-the-art performance by learning from random images

New techniques

Performance and future work

VentureBeat

Why machine learning strategies fail

Machine learning is about data

AI talent is still in high demand

Outsourcing AI talent

How do you evaluate your AI strategy?

VentureBeat

How to know if federated learning should be part of your data strategy

VentureBeat

Recent Posts

Categories

Archives