Machine learning has been applied to many health-related tasks, such as the development of new medical treatments, the management of patient data and records, and the treatment of chronic diseases. To achieve success in those SOTA applications, we must rely on the time-consuming technique of model building evaluation. To alleviate this load, Yue Zhao et al have proposed a PyHealth, a Python-based toolbox. As the name implies, this toolbox contains a variety of ML models and architecture algorithms for working with medical data. In this article, we will go through this model to understand its working and application. Below are the major points that we are going to discuss in this article.

Lets first discuss the use case of machine learning in the healthcare industry.

Machine learning is being used in a variety of healthcare settings, from case management of common chronic conditions to leveraging patient health data in conjunction with environmental factors such as pollution exposure and weather.

Machine learning technology can assist healthcare practitioners in developing accurate medication treatments tailored to individual features by crunching enormous amounts of data. The following are some examples of applications that can be addressed in this segment:

The ability to swiftly and properly diagnose diseases is one of the most critical aspects of a successful healthcare organization. In high-need areas like cancer diagnosis and therapy, where hundreds of drugs are now in clinical trials, scientists and computationalists are entering the mix. One method combines cognitive computing with genetic tumour sequencing, while another makes use of machine learning to provide diagnosis and treatment in a range of fields, including oncology.

Medical imaging, and its ability to provide a complete picture of an illness, is another important aspect in diagnosing an illness. Deep learning is becoming more accessible as data sources become more diverse, and it may be used in the diagnostic process, therefore it is becoming increasingly important. Although these machine learning applications are frequently correct, they have some limitations in that they cannot explain how they came to their conclusions.

ML has the potential to identify new medications with significant economic benefits for pharmaceutical companies, hospitals, and patients. Some of the worlds largest technology companies, like IBM and Google, have developed ML systems to help patients find new treatment options. Precision medicine is a significant phrase in this area since it entails understanding mechanisms underlying complex disorders and developing alternative therapeutic pathways.

Because of the high-risk nature of surgeries, we will always need human assistance, but machine learning has proved extremely helpful in the robotic surgery sector. The da Vinci robot, which allows surgeons to operate robotic arms in order to do surgery with great detail and in confined areas, is one of the most popular breakthroughs in the profession.

These hands are generally more accurate and steady than human hands. There are additional instruments that employ computer vision and machine learning to determine the distances between various body parts so that surgery can be performed properly.

Health data is typically noisy, complicated, and heterogeneous, resulting in a diverse set of healthcare modelling issues. For instance, health risk prediction is based on sequential patient data, disease diagnosis based on medical images, and risk detection based on continuous physiological signals.

Electroencephalogram (EEG) or electrocardiogram (ECG), for example, and multimodal clinical notes (e.g., text and images). Despite their importance in healthcare research and clinical decision making, the complexity and variability of health data and tasks need the long-overdue development of a specialized ML system for benchmarking predictive health models.

PyHealth is made up of three modules: data preprocessing, predictive modelling, and assessment. Both computer scientists and healthcare data scientists are PyHealths target consumers. They can run complicated machine learning processes on healthcare datasets in less than 10 lines of code using PyHealth.

The data preprocessing module converts complicated healthcare datasets such as longitudinal electronic health records, medical pictures, continuous signals (e.g., electrocardiograms), and clinical notes into machine learning-friendly formats.

The predictive modelling module offers over 30 machine learning models, including known ensemble trees and deep neural network-based approaches, using a uniform yet flexible API geared for both researchers and practitioners.

The evaluation module includes a number of evaluation methodologies (for example, cross-validation and train-validation-test split) as well as prediction model metrics.

There are five distinct advantages to using PyHealth. For starters, it contains more than 30 cutting-edge predictive health algorithms, including both traditional techniques like XGBoost and more recent deep learning architectures like autoencoders, convolutional based, and adversarial based models.

Second, PyHealth has a broad scope and includes models for a variety of data types, including sequence, image, physiological signal, and unstructured text data. Third, for clarity and ease of use, PyHealth includes a unified API, detailed documentation, and interactive examples for all algorithmscomplex deep learning models can be implemented in less than ten lines of code.

Fourth, unit testing with cross-platform, continuous integration, code coverage, and code maintainability checks are performed on most models in PyHealth. Finally, for efficiency and scalability, parallelization is enabled in select modules (data preprocessing), as well as fast GPU computation for deep learning models via PyTorch.

PyHealth is a Python 3 application that uses NumPy, scipy, scikit-learn, and PyTorch. As shown in the diagram below, PyHealth consists of three major modules: First is the data preprocessing module can validate and convert user input into a format that learning models can understand;

Second is the predictive modelling module is made up of a collection of models organized by input data type into sequences, images, EEG, and text. For each data type, a set of dedicated learning models has been implemented, and the third is the evaluation module can automatically infer the task type, such as multi-classification, and conduct a comprehensive evaluation by task type.

Most learning models share the same interface and are inspired by the scikit-API learn to design and general deep learning design: I fit learns the weights and saves the necessary statistics from the train and validation data; load model chooses the model with the best validation accuracy, and inference predicts the incoming test data.

For quick data and model exploration, the framework includes a library of helper and utility functions (check parameter, label check, and partition estimators). For example, a label check can check the data label and infer the task type, such as binary classification or multi-classification, automatically.

PyHealth for model building

Now below well discuss how we can leverage the API of this framework. First, we need to install the package by using pip.

! pip install pyhealth

Next, we can load the data from the repository itself. For that, we need to clone the repository. After cloning the repository inside the datasets folder there is a variety of datasets like sequenced based, image-based, etc. We are using the mimic dataset and it is in the zip form we need to unzip it. Below is the snippet clone repository, and unzip the data.

The unzipped file is saved in the current working directory with the name of the folder as a mimic. Next to use this dataset we need to load the sequence data generator function which serves as functionality to prepare the dataset for experimentation.

Now we have loaded the dataset. Now we can do further modelling as below.

Here is the fitment result.

Through this article, we have discussed how machine learning can be used in the healthcare industry by observing the various applications. As this domain is being quite vast and N number application, we have discussed a Python-based toolbox that is designed to build a predictive modelling approach by using various deep learning techniques such as LSTM, GRU for sequence data, and CNN for image-based data.

Read the original:
How to build healthcare predictive models using PyHealth? - Analytics India Magazine

This article is part of our series that explores thebusiness of artificial intelligence

Digital technologies, and at their forefront artificial intelligence, are triggering fundamental shifts in society, politics, education, economy, and other fundamental aspects of life. These changes provide opportunities for unprecedented growth across different sectors of the economy. But at the same time, they entail challenges that organizations must overcome before they can tap into their full potential.

In a recent talk at an online conference organized by Stanford Human-Centered Artificial Intelligence (HAI), Stanford professor Erik Brynjolfsson discussed some of these opportunities and challenges.

Brynjolfsson, who directs Stanfords Digital Economy Lab, believes that in the coming decade, the use of artificial intelligence will be much more widespread than it is today. But its adoption will also face a period of lull, also known as the J-curve.

Theres a growing gap between what the technology is capable of and what it is already doing versus how we are responding to that, Brynjolfsson says. And thats where a lot of our societys biggest challenges and problems and some of our biggest opportunities lie.

According to Brynjolfsson, the next decade will see significantly higher productivity thanks to a wave of powerful technologiesespecially machine learningthat are finding their way into every computing device and application.

Advances in computer vision have been tremendous, especially in areas such as image recognition and medical imaging. Talking to phones, watches, and smart speakers has become commonplace thanks to advances in natural language processing and speech recognition. Product recommendation, ad placement, insurance underwriting, loan approval, and many other applications have benefited immensely from advances in machine learning.

In many areas, machine learning is reducing costs and accelerating production. For example, the application of large language models in programming can help software developers become much more productive and achieve more in less time.

In other areas, machine learning can help create applications that did not exist before. For example, generative deep learning models are creating new applications for arts, music, and other creative work. In areas such as online shopping, advances in machine learning can create major shifts in business models, such as moving from shopping-then-shipping to shipping-then-shopping.

The lockdowns and urgency caused by the covid-19 pandemic accelerated the adoption of these technologies in different sectors, including remote work tools, robotic process automation, powered drug research, and factory automation.

The pandemic has been horrific in so many ways, but another thing its done is its accelerated the digitization of the economy, compressing in about 20 weeks what would have taken maybe 20 years of digitization, Brynjolfsson says. Weve all invested in technologies that are allowing us to adapt to a more digital world. Were not going to stay as remote as we are now, but were not going all the way back either. And that increased digitization of business processes and skills compresses the timeframe for us to adopt these new ways of working and ultimately drive higher productivity.

The productivity potential of machine learning technologies has one big caveat.

Historically, when these new technologies become available, they dont immediately translate into productivity growth. Often theres a period where productivity declines, where theres a lull, Brynjolfsson says. And the reason theres this lull is that you need to reinvent your organizations, you need to develop new business processes.

Brynjolfsson calls this the Productivity J-Curve and has documented it in a paper published in the American Economic Journal: Macroeconomics. Basically, the great potential caused by new general-purpose technologies like the steam engine, electricity, and more recently machine learning requires fundamental changes in business processes and workflows, the co-invention of new products and business models, and investment in human capital.

These investments and changes often take several years, and during this period, they dont yield tangible results. During this phase, the companies are creating intangible assets, according to Brynjolfsson. For example, they might be training and reskilling their workforce to employ these new technologies. They might be redesigning their factories or instrumenting them with new sensor technologies to take advantage of machine learning models. They might need to revamp their data infrastructure and create data lakes on which they can train and run ML models.

These efforts might cost millions of dollars (or billions in the case of large corporations) and make no change in the companys output in the short term. At first glance, it seems that costs are increasing without any return on investment. When these changes reach their turning point, they result in a sudden increase in productivity.

Were in this period right now where were making a lot of that painful transition, restructuring work, and theres a lot of companies that are struggling with that, Brynjolfsson says. But were working through that, and these J-curves will lead to higher productivityaccording to our research, were near the bottom and turning up.

Unfortunately, adapting to AI and other new digital technologies does not run on a predictable path. Most firms arent making the transition correctly or lack the creativity and understanding to make the transition. Various studies show that most applied machine learning projects fail.

Only about the top 10-15 percent of firms are doing most of the investment in these intangibles. The other 85-90 percent of firms are lagging behind and are hardly making any of these restructuring needed, Brynjolfsson says. This is not just the big tech firms. This is within every industry, manufacturing, retail, finance, resources. In each category, were seeing the leading firms pulling away from the rest. Theres a growing performance gap.

But while adopting new technologies is going to be difficult, it is happening at a much faster pace in comparison to previous cycles of technological advances because we are better prepared to make the transition.

I think what is becoming clear is that its going to happen a lot faster in part because we have a much more professional class of people trying to study what works and what doesnt work, Brynjolfsson says. Some of them are in business schools and academia. A lot of them are in consulting companies. Some of them are journalists. And there are people who are describing which practices work and which dont.

Another element that can help immensely is the availability of machine learning and data science tools to process and study the huge amounts of data available on organizations, people, and the economy.

For example, Brynjolfsson and his colleagues are working on a big dataset of 200 million job postings, which include the full text of the job description along with other information. Using different machine learning models and natural language processing techniques, they can transform the job posts into numerical vectors that can then be used for various tasks.

We think of all the jobs as this mathematical space. We can understand how they can relate to each other, Brynjolfsson says.

For example, they can make simple inferences such as how similar or different two or more job posts are based on their text descriptions. They can use other techniques such as clustering and graph neural networks to draw more important conclusions such as what kind of skills are more in demand, or how would the characteristics of a job post change if you modified the description to add AI skills such as Python or TensorFlow. Companies can use these models to find holes in their hiring strategies or to analyze the hiring decisions of their competitors and leading organizations.

Those kinds of tools just didnt exist as recently as five years ago, and I think its a revolution that is just as important as the microscope or some of the other revolutions in science, Brynjolfsson says. We now have them for social sciences and business to have this kind of visibility. Thats allowing us to make a transition a lot more rapidly than before.

However, Brynjolfsson warns that not many companies are using these kinds of tools. This is perhaps further testament to his previous point that companies have not yet figured out the right transition strategy and are relying on old methods to restructure and adapt themselves to the age of AI. And at the center of this strategy should be the correct use of human capital.

You have hundreds of billions of dollars of human capital, of skills walking out the door, and then the company tries to hire back people with the skills that they need. What they dont realize is that the workers that they let go often had skills that were very adjacent to the ones theyre hiring for, Brynjolfsson says.

With the help of machine learning, they will have better visibility and knowledge of their skill adjacencies, Brynjolfsson says. For example, a company might discover that instead of laying off a bunch of people and looking to hire new talent, perhaps all they need to do is a little bit of retraining and repurposing of their workforce.

Its much more expensive to hire somebody fresh than would have been for them to take some of those people who are already in the company and say, if we teach you Python or customer service skills or other skills, you can be doing this job that were looking to hire people for, Brynjolfsson says. My hope is that, in the coming decade, workers will be in a much better position to take full advantage of their capabilities and skills. And it will be good for the companies too to understand all the assets that they have in there, and machine learning can help a lot with understanding those relationships.

Excerpt from:
AIs J-curve and upcoming productivity boom - TechTalks

AUSTIN, Texas, Jan. 24, 2022 /PRNewswire/ -- Silicon Labs, a leader in secure, intelligent wireless technology for a more connected world, today announced the BG24 and MG24 families of 2.4 GHz wireless SoCs for Bluetooth and Multiple-protocol operations, respectively, and a new software toolkit. This new co-optimized hardware and software platform will help bring AI/ML applications and wireless high performance to battery-powered edge devices. Matter-ready, the ultra-low-power BG24 and MG24 families support multiple wireless protocols and incorporate PSA Level 3 Secure Vaultprotection, ideal for diverse smart home, medical and industrial applications. The SoC and software solution for the Internet of Things (IoT) announced today includes:

"The BG24 and MG24 wireless SoCs represent an awesome combination of industry capabilities including broad wireless multiprotocol support, battery life, machine learning, and security for IoT Edge applications," said Matt Johnson, CEO of Silicon Labs.

First Integrated AI/ML Acceleration Improves Performance and Energy Efficiency

IoT product designers see the tremendous potential of AI and machine learning to bring even greater intelligence to edge applications like home security systems, wearable medical monitors, sensors monitoring commercial facilities and industrial equipment, and more. But today, those considering deploying AI or machine learning at the edge are faced with steep penalties in performance and energy use that may outweigh the benefits.

The BG24 and MG24 alleviate those penalties as the first ultra-low powered devices with dedicated AI/ML accelerators built-in. This specialized hardware is designed to handle complex calculations quickly and efficiently, with internal testing showing up to a 4x improvement in performance along with up to a 6x improvement in energy efficiency. Because the ML calculations are happening on the local device rather than in the cloud, network latency is eliminated for faster decision-making and actions.

The BG24 and MG24 families also have the largest Flash and random access memory (RAM) capacities in the Silicon Labs portfolio. This means that the device can evolve for multi-protocol support, Matter, and trained ML algorithms for large datasets. PSA Level3-Certified Secure VaultTM,the highest level of security certification for IoT devices, provides the security needed in products like door locks, medical equipment, and other sensitive deployments where hardening the device from external threats is paramount.

To learn more about the capabilities of the BG24 and MG24 SoCs and view a demo on how to get started, register for the instructional Tech Talk "Unboxing the new BG24 and MG24 SoCs" here: https://www.silabs.com/tech-talks.

AI/ML Software and Matter-Support Help Designers Create for New Innovative Applications

In addition to natively supporting TensorFlow, Silicon Labs has partnered with some of the leading AI and ML tools providers, like SensiML and Edge Impulse, to ensure that developers have an end-to-end toolchain that simplifies the development of machine learning models optimized for embedded deployments of wireless applications. Using this new AI/ML toolchain with Silicon Labs's Simplicity Studio and the BG24 and MG24 families of SoCs, developers can create applications that draw information from various connected devices, all communicating with each other using Matter to then make intelligent machine learning-driven decisions.

For example, in a commercial office building, many lights are controlled by motion detectors that monitor occupancy to determine if the lights should be on or off. However, when typing at a desk with motion limited to hands and fingers, workers may be left in the dark when motion sensors alone cannot recognize their presence. By connecting audio sensors with motion detectors through the Matter application layer, the additional audio data, such as the sound of typing, can be run through a machine-learning algorithm to allow the lighting system to make a more informed decision about whether the lights should be on or off.

ML computing at the edge enables other intelligent industrial and home applications, including sensor-data processing for anomaly detection, predictive maintenance, audio pattern recognition for improved glass-break detection, simple-command word recognition, and vision use cases like presence detection or people counting with low-resolution cameras.

Alpha Program Highlights Variety of Deployment Options

More than 40 companies representing various industries and applications have already begun developing and testing this new platform solution in a closed Alpha program. These companies have been drawn to the BG24 and MG24 platforms by their ultra-low power, advanced features, including AI/ML capabilities and support for Matter. Global retailers are looking to improve the in-store shopping experience with more accurate asset tracking, real-time price updating, and other uses. Participants from the commercial building management sector are exploring how to make their building systems, including lighting and HVAC, more intelligent to lower owners' costs and reduce their environmental footprint. Finally, consumer and smart home solution providers are working to make it easier to connect various devices and expand the way they interact to bring innovative new features and services to consumers.

Silicon Labs' Most Capable Family of SoCs

The single-die BG24 and MG24 SoCs combine a 78 MHz ARM Cortex-M33 processor, high-performance 2.4 GHz radio, industry-leading 20-bit ADC, an optimized combination of Flash (up to 1536 kB) and RAM (up to 256 kB), and an AI/ML hardware accelerator for processing machine learning algorithms while offloading the ARM Cortex-M33, so applications have more cycles to do other work.Supporting a broad range of 2.4 GHz wireless IoT protocols, these SoCs incorporate the highest security with the best RF performance/energy-efficiency ratio in the market.

Availability

EFR32BG24 and EFR32MG24 SoCs in 5 mm x 5 mm QFN40 and 6 mm x 6 mm QFN48 packages are shipping today to Alpha customers and will be available for mass deployment in April 2022. Multiple evaluation boards are available to designers developing applications.Modules based on the BG24 and MG24 SoCs will be available in the second half of 2022.

To learn more about the new BG24 family, go to: http://silabs.com/bg24.

To learn more about the new MG24 family, go to: http://silabs.com/mg24.

To learn more about how Silicon Labs supports AI and ML, go to: http://silabs.com/ai-ml.

About Silicon Labs

Silicon Labs (NASDAQ: SLAB) is a leader in secure, intelligent wireless technology for a more connected world. Our integrated hardware and software platform, intuitive development tools, unmatched ecosystem, and robust support make us an ideal long-term partner in building advanced industrial, commercial, home, and life applications. We make it easy for developers to solve complex wireless challenges throughout the product lifecycle and get to market quickly with innovative solutions that transform industries, grow economies, and improve lives.Silabs.com

View original content to download multimedia:https://www.prnewswire.com/news-releases/silicon-labs-brings-ai-and-machine-learning-to-the-edge-with-matter-ready-platform-301466032.html

SOURCE Silicon Labs

Read the original post:
Silicon Labs Brings AI and Machine Learning to the Edge with Matter-Ready Platform - inForney.com

There is great promise and potential in artificial intelligence (AI), but if such technologies are built and trained by humans, are they capable of bias?

Absolutely, says William Wang, the Duncan and Suzanne Mellichamp Chair in Artificial Intelligence and Designs at UC Santa Barbara, who will give the virtual talk What is Responsible AI, at 4 p.m. Tuesday, Jan. 25, as part of the UCSB Librarys Pacific Views speaker series (register here).

The key challenge for building AI and machine learning systems is that when such asystem is trained on datasets with limited samples from history, they may gain knowledge from the protected variables (e.g., gender, race, income, etc.), and they are prone to produce biased outputs, said Wang, also director of UC Santa Barbaras Center for Responsible Machine Learning.

Sometimes these biases could lead to the rich getting richer phenomenon after the AI systems are deployed, he added.Thats why in addition to accuracy, it is important to conduct research in fair and responsible AI systems, including the definition of fairness, measurement, detection and mitigation of biases in AI systems.

Wangs examination of the topic serves as the kickoff event for UCSB Reads 2022, the campus and community-wide reading program run by UCSB Library. Their new season is centered on Ted Chiangs Exhalation, a short story collection thataddressesessential questions about human and computer interaction, including the use of artificial intelligence.

Copies of Exhalation will be distributed free to students (while supplies last) Tuesday, Feb. 1 outside the Librarys WestPaseo entrance. Additional events announced so far include on-air readings from the book on KCSB, a faculty book discussion moderated by physicist and professor David Weld and a sci-fi writing workshop. It all culminates May 10 with a free lecture by Ted Chiang in Campbell Hall.

First though: William Wang, an associate professor of computer science and co-director of the Natural Language Processing Group.

In this talk, my hope is to summarize the key advances of artificialintelligence technologies in the last decade, and share how AI can bring us an exciting future, he noted. I will also describe the key challenges of AI: how we should consider the research and development of responsible AI systems,which not only optimize their accuracy performance,but also provide a human-centric view to consider fairness, bias, transparency and energy efficiency of AI systems.

How do we build AI models that are transparent? How do we write AI system descriptions that meet disclosive transparency guidelines?How do we consider energy efficiency when building AI models? he asked. The future of AI is bright, but all of these are key aspects of responsible AI that we need to address.

See more here:
Getting a Read on Responsible AI | The UCSB Current - The UCSB Current