A few years ago, I wondered how edge data would ever be useful given the enormous cost of transmitting all the data to either the centralized data center or some variant of cloud infrastructure. (It is said that 5G will solve that problem).

Consider, for example, applications of vast sensor networks that stream a great deal of data at small intervals. Vehicles on the move are a good example.

There is telemetry from cameras, radar, sonar, GPS and LIDAR, the latter about 70MB/sec. This could quickly amount to four terabytes per day (per vehicle). How much of this data needs to be retained? Answers I heard a few years ago were along two lines:

My counterarguments at the time were:

Introducing TensorFlow federated, via The TensorFlow Blog:

This centralized approach can be problematic if the data is sensitive or expensive to centralize. Wouldn't it be better if we could run the data analysis and machine learning right on the devices where that data is generated, and still be able to aggregate together what's been learned?

Since I looked at this a few years ago, the distinction between an edge device and a sensor has more or less disappeared. Sensors can transmit via wifi (though there is an issue of battery life, and if they're remote, that's a problem); the definition of the edge has widened quite a bit.

Decentralized data collection and processing have become more powerful and able to do an impressive amount of computing. The case is point in Intel's Introducing the Intel Neural Compute Stick 2 computer vision and deep learning accelerator powered by the Intel Movidius Myriad X VPU, that can stick into a Pi for less than $70.00.

But for truly distributed processing, the Apple A13 chipset in the iPhone 11 has a few features that boggle the mind: From Inside Apple's A13 Bionic system-on-chip Neural Engine, a custom block of silicon separate from the CPU and GPU, focused on accelerating Machine Learning computations. The CPU has a set of "machine learning accelerators" that perform matrix multiplication operations up to six times faster than the CPU alone. It's not clear how exactly this hardware is accessed, but for tasks like machine learning (ML) that use lots of matrix operations, the CPU is a powerhouse. Note that this matrix multiplication hardware is part of the CPU cores and separate from the Neural Engine hardware.

This should beg the question, "Why would a smartphone have neural net and machine learning capabilities, and does that have anything to do with the data transmission problem for the edge?" A few years ago, I thought the idea wasn't feasible, but the capability of distributed devices has accelerated. How far-fetched is this?

Let's roll the clock back thirty years. The finance department of a large diversified organization would prepare in the fall a package of spreadsheets for every part of the organization that had budget authority. The sheets would start with low-level detail, official assumptions, etc. until they all rolled up to a small number of summary sheets that were submitted headquarters. This was a terrible, cumbersome way of doing things, but it does, in a way, presage the concept of federated learning.

Another idea that vanished is Push Technology that shared the same network load as centralizing sensor data, just in the opposite direction. About twenty-five years, when everyone had a networked PC on their desk, the PointCast Network used push technology. Still, it did not perform as well as expected, often believed to be because its traffic burdened corporate networks with excessive bandwidth use, and was banned in many places. If Federated Learning works, those problems have to be addressed

Though this estimate changes every day, there are 3 billion smartphones in the world and 7 billion connected devices.You can almost hear the buzz in the air of all of that data that is always flying around. The canonical image of ML is that all of that data needs to find a home somewhere so that algorithms can crunch through it to yield insights. There are a few problems with this, especially if the data is coming from personal devices, such as smartphones, Fitbit's, even smart homes.

Moving highly personal data across the network raises privacy issues. It is also costly to centralize this data at scale. Storage in the cloud is asymptotically approaching zero in cost, but the transmission costs are not. That includes both local WiFi from the devices (or even cellular) and the long-distance transmission from the local collectors to the central repository. This s all very expensive at this scale.

Suppose, large-scale AI training could be done on each device, bringing the algorithm to the data, rather than vice-versa? It would be possible for each device to contribute to a broader application while not having to send their data over the network. This idea has become respectable enough that it has a name - Federated Learning.

Jumping ahead, there is no controversy that training a network without compromising device performance and user experience, or compressing a model and resorting to a lower accuracy are not alternatives. In Federated Learning: The Future of Distributed Machine Learning:

To train a machine learning model, traditional machine learning adopts a centralized approach that requires the training data to be aggregated on a single machine or in a datacenter. This is practically what giant AI companies such as Google, Facebook, and Amazon have been doing over the years. This centralized training approach, however, is privacy-intrusive, especially for mobile phone usersTo train or obtain a better machine learning model under such a centralized training approach, mobile phone users have to trade their privacy by sending their personal data stored inside phones to the clouds owned by the AI companies.

The federated learning approach decentralizes training across mobile phones dispersed across geography. The presumption is that they collaboratively develop machine learning while keeping their personal data on their phones. For example, building a general-purpose recommendation engine for music listeners. While the personal data and personal information are retained on the phone, I am not at all comfortable that data contained in the result sent to the collector cannot be reverse-engineered - and I havent heard a convincing argument to the contrary.

Here is how it works. A computing group, for example, is a collection of mobile devices that have opted to be part of a large scale AI program. The device is "pushed" a model and executes it locally and learns as the model processes the data. There are some alternatives to this. Homogeneous models imply that every device is working with the same schema of data. Alternatively, there are heterogeneous models where harmonization of the data happens in the cloud.

Here are some questions in my mind.

Here is the fuzzy part: federated learning sends the results of the learning as well as some operational detail such as model parameters and corresponding weights back to the cloud. How does it do that and preserve your privacy and not clog up your network? The answer is that the results are a fraction of the data, and since the data itself is not more than a few Gb, that seems plausible. The results sent to the cloud can be encrypted with, for example, homomorphic encryption (HE). An alternative is to send the data as a tensor, which is not encrypted because it is not understandable by anything but the algorithm. The update is then aggregated with other user updates to improve the shared model. Most importantly, all the training data remains on the user's devices.

In CDO Review, The Future of AI. May Be In Federated Learning:

Federated Learning allows for faster deployment and testing of smarter models, lower latency, and less power consumption, all while ensuring privacy. Also, in addition to providing an update to the shared model, the improved (local) model on your phone can be used immediately, powering experiences personalized by the way you use your phone.

There is a lot more to say about this. The privacy claims are a little hard to believe. When an algorithm is pushed to your phone, it is easy to imagine how this can backfire. Even the tensor representation can create a problem. Indirect reference to real data may be secure, but patterns across an extensive collection can surely emerge.

Here is the original post:
Federated machine learning is coming - here's the questions we should be asking - Diginomica

In the last two decades, Artificial Intelligence has steadily made its way into versatile industry applications. This has helped businesses reap major rewards by reducing operational costs, triggering efficiency, boosting revenue and improving the overall customer experience. With a constantly evolving range of technologies, efforts are on to develop AI to a stage where it reduces human intervention to the minimum. This is where the relevance of Machine Learning and Cloud comes in. As businesses transform the way in which they communicate, work and grow, the importance of Cloud in deploying Machine Learning models becomes important. Because of the massive storage and processing of data, Machine Learning often involves the application of computational power to train models because of the lack of strong machines. Thus, when Cloud is paired with ML models, it forms the Intelligent Cloud that becomes a suitable destination for any companys Machine Learning projects. The Cloud will enable ML data to make more accurate predictions and analyze data more efficiently, enhancing business value by a huge extent. With so many developments, any study of Machine Learning is incomplete without learning about its association with the Cloud.To help working professionals become a part of any companys end-to-end packaged ML solution, IIT Madras in collaboration with upGrad, has launched the ML and Cloud program. As one of Indias largest online education platforms, it recognizes the huge potential of taking Machine Learning to the Cloud, and how the first step to enable this is to train ML professionals in the right direction. Lets take a look at how upGrads ML program in Cloud will help professionals skill up in the foreseeable future.'; var randomNumber = Math.random(); var isIndia = (window.geoinfo && window.geoinfo.CountryCode === 'IN') && (window.location.href.indexOf('outsideindia') === -1 ); console.log(isIndia && randomNumber A revolutionary advanced certification course in Machine Learning and CloudIn the current business set up, data and insights can be termed as the true currency for business operation. This is why every organization is immensely scaling up its ML capabilities. upGrads Advanced Certification in Machine Learning and Cloud is helping learners become Machine Learning experts by training them to deploy machine learning models using PySpark on Cloud. This prestigious certification provides students with the opportunity to learn from a set of experienced Machine Learning faculty and industry leaders. Another highlight of this 9-month program is that it has about 300+ hiring partners, ensuring that professionals who choose to upskill with this course ends up in the industry of their choice.

The Advanced Certification in Machine Learning and Cloud by upGrad seeks to build employability of professionals and also boost up their annual packages. The requirement for ML professionals has now percolated to multiple industry domains like e-commerce, retail, healthcare, banking, manufacturing, transport, NBFC, and finance, among others. The course offers an equal opportunity to every learner, enhancing their relevance in the company they will work for. In Data and ML related hirings, recruiters look for people who are proficient and knowledgeable and can prove to be assets to employers in the company. This certification program by upGrad is an excellent opportunity to make a credible career transition. Considering ML is one of the fastest-growing fields in Data today, Machine Learning engineers are getting hired at astounding pay packages. In fact, an Indeed survey revealed that there has been more than a 300% spike in ML hirings since 2015. Considering this shift, upGrads Advanced Program in Machine Learning and Cloud is the best way to flag off ones ML journey.

Top skills that the program will offer Programming: Learners will be working in core and necessary languages like Python and SQL since the former is required for ML and the latter for the Cloud.

Machine learning concepts: The program is set to offer a holistic understanding of both basic and advanced subjects within ML. This includes the application of the appropriate ML algorithm to categorize unknown data or make predictions about it. Also included is the ability to modify and craft algorithms of your own should and when the need arises.

Foundations of Cloud and Hadoop: It also included knowledge about Hadoop, Hive, and HDFS along with the implementation of ML algorithms in the cloud on Spark/ PySpark (AWS/ Azure/ GCP). Overall, the curriculum is designed so that students learn the local Python implementation as well as the cloud PySpark implementation of classical machine learning algorithms.

Who should apply for this program?Keeping the overall market landscape in mind, this program by upGrad is ideal for the following categories:

The pedagogy and content of upGrads Advanced Program in Machine Learning and Cloud is a perfect integration of online lectures, offline engagement, practical case studies, and interactive networking sessions. The platform provides full support to young professionals in their ML journey by also catering to the needs of employers by training the future workforce in all data-related aspects. Whether it is resume feedback, mock interview sessions with industry experts, or conducting placement drives with top-notch companies, upGrad has provided it all to its learners. Many of these learners have also been placed at companies like KPMG, Uber, Big Basket, Bain & Co, Pwc, Zivame, Fractal Analytics, Microsoft etc. with impressive salary shifts.

See original here:
Looking for an impressive salary hike? Power up your career with upGrads Machine Learning and Cloud prog - Times of India

In 2020, small to medium sized businesses (SMBs) are likely to focus more on supporting workers to travel and collaborate in ways that suit them, while still facing a clear economic imperative to keep costs under control.

This will likely involve increased use of technologies such as machine learning and automation to: help determine and enforce spending policies; ensure people travelling for work can optimise, track, and analyse their spend; and prioritise travel options that meet goals around environmental responsibility and sustainability.

Businesses that recognise and respond to these trends will be better-placed to save money while improving employee engagement and performance, according to SAP Concur.

Fabian Calle, General Manager, Small to Medium Business, ANZ, SAP Concur, said, As the new decade begins, the business environment will be subject to the same economic ups and downs seen in the previous decade. However, with new technologies and approaches, most businesses will be able to leverage automation and even artificial intelligence to smooth out those peaks and troughs.

SAP Concur has identified the top five 2020 predictions for SMBs, covering economics, technology, business, travel, the environment, diversity, and corporate social responsibility:

Calle said, 2020 will continue to drive significant developments as organisations of all sizes look to optimise efficiency and productivity through employee operations and satisfaction. Australian businesses need to be aware of these trends and adopt cutting edge technology to facilitate their workers need to travel and collaborate more effectively and with less effort.

Read more from the original source:
Machine learning and eco-consciousness key business trends in 2020 - Finfeed

Deep learning offers the promise of bypassing the procedure of manual feature engineering by learning representations in conjunction with statistical models in an end-to-end fashion. In any case, neural network architectures themselves are ordinarily designed by specialists in a painstaking, ad hoc fashion. Neural architecture search (NAS) has been touted as the way ahead for lightening this agony via automatically identifying architectures that are better than hand-planned ones.

Machine learning has given some huge achievements in diverse fields as of late. Areas like financial services, healthcare, retail, transportation, and more have been utilizing machine learning frameworks somehow, and the outcomes have been promising.

Machine learning today isnt constrained to R&D applications however, has made its foray into the enterprise space. However, the conventional ML process is human-dependent, and not all companies have the assets to put resources into an experienced data science team. AutoML might be the answer to such circumstances.

AutoML focuses on automating each part of the machine learning (ML) work process to increase effectiveness and democratize machine learning so that non-specialists can apply machine learning to their issues effortlessly. While AutoML includes the automation of a wide scope of problems related with ETL (extract, transform, load), model training, and model development, the issue of hyperparameter enhancement is a core focus of AutoML. This issue includes configuring the internal settings that govern the conduct of an ML model/algorithm so as to restore a top-notch prescient model.

Creating neural network models frequently requires noteworthy architecture engineering. You can sometimes get by with transfer learning, yet if you truly need the most ideal performance its generally best to structure your very own network. This requires particular skills(read: costly from a business point of view) and is challenging in general; we may not know the cutoff points of the present cutting edge methods! Its a ton of experimentation and the experimentation itself is tedious and costly.

The NAS discovered architecture is trained and tried on a lot of smaller-than-real world dataset. This is done in light of the fact that training on something enormous, like ImageNet, would take an extremely significant time-frame. In any case, the thought is that a network that performs better on a smaller, yet comparatively organized dataset should likewise perform better on a bigger and progressively complex one, which has commonly been valid in the deep learning time.

Second, is that the search space itself is very constrained. NAS is intended to construct architectures that are fundamentally the same as in style to the current state-of-the-art. For image recognition, this is to have a set of repeated blocks in the network while continuously downsampling. The set of blocks to browse to manufacture the rehashing ones are additionally usually utilized in current research. The principal novel part of the NAS discovered networks is the manner by which the blocks are connected together.

The demand for machine learning systems has taken off in the course of recent years. This is because of the achievement of ML in a wide range of applications today. Nonetheless, even with this unmistakable sign, that machine learning can give lifts to specific organizations, a lot of organizations struggle to deploy ML models.

To start with, they have to set up a team of seasoned data scientists who order a top-notch pay. Second, regardless of whether you have an extraordinary team, choosing which model is the best for your concern frequently requires more experience than information. The achievement of machine learning in a wide scope of applications has led to a consistently developing demand for machine learning frameworks that can be utilized off the rack by non-experts. AutoML will, in general, automate the greatest number of steps in an ML pipeline, with a minimum amount of human effort and without trading off the models performance.

Argonne analysts have made a neural architecture search that automates the development of deep learning-based predictive models for cancer data. While expanding swaths of collected information and growing sizes of computing power are assisting with improving our comprehension of cancer, further improvement of data-driven strategies for the diseases diagnosis, detection and prognosis are necessary. There is a specific need to grow deep learning techniques -; that is, machine learning algorithms equipped for extracting science from unstructured information.

Analysts from the U.S. Division of Energys (DOE) Argonne National Laboratory have made progress toward accelerating such efforts by exhibiting a strategy for the automated generation of neural networks.

Architecture search has become unmistakably increasingly proficient; finding a network with a single GPU in a single day of training as with ENAS is quite astonishing. In any case, our search space is still actually very constrained. The present NAS algorithms despite everything utilize the structures and building blocks that were hand-planned, they simply set up them together in an unexpected way!

A solid and conceivably groundbreaking future direction would be a far more extensive-ranging search, to truly search for novel architectures. Such algorithms may uncover significantly increasingly hidden deep learning insider facts within these huge and complex systems. Obviously, such search space requires efficient algorithm design. This new bearing of NAS and AutoML gives exciting challenges to the AI community, and actually a possibility for another breakthrough in the science.

Read the original:
Neural Architecture and AutoML Technology - Analytics Insight