Technology
Researchers Develop Algorithm for Optimal Decision Making Under Heavy-tailed Noisy Rewards
By Adedapo Adesanya
The exploration algorithms for stochastic multi-armed bandits (MABs)–sequential decision-making problems under uncertain environments–typically assume light-tailed distributions for reward noises.
However, real-world datasets often show heavy-tailed noise. In light of this, researchers from Korea propose an algorithm that can achieve minimax optimality (minimum loss under maximum loss scenario) with minimal prior information.
Superior to existing algorithms, the new algorithm has potential applications in autonomous trading and personalized recommendation systems.
In data science, researchers typically deal with data that contain noisy observations. An important problem explored by data scientists in this context is the problem of sequential decision-making. This is commonly known as a “stochastic multi-armed bandit” or (stochastic MAB).
Here, an intelligent agent sequentially explores and selects actions based on noisy rewards under an uncertain environment. Its goal is to minimize cumulative regret–the difference between the maximum reward and the expected reward of selected actions. A smaller regret implies more efficient decision-making.
Most existing studies on stochastic MABs have performed regret analysis under the assumption that the reward noise follows a light-tailed distribution. However, many real-world datasets, in fact, show a heavy-tailed noise distribution.
These include user behavioural pattern data used for developing personalized recommendation systems, stock price data for automatic transaction development, and sensor data for autonomous driving.
In a recent study, Assistant Professor Kyungjae Lee of Chung-Ang University and Assistant Professor Sungbin Lim of the Ulsan Institute of Science and Technology, both in Korea, addressed this issue. In their theoretical analysis, they proved that the existing algorithms for stochastic MABs were sub-optimal for heavy-tailed rewards.
More specifically, the methods employed in these algorithms–robust upper confidence bound (UCB) and adaptively perturbed exploration (APE) with unbounded perturbation–do not guarantee a minimax (minimization of maximum possible loss) optimality.
“Based on this analysis, minimax optimal robust (MR) UCB and APE methods have been proposed. MR-UCB utilizes a tighter confidence bound of robust mean estimators, and MR-APE is its randomized version. It employs bounded perturbation whose scale follows the modified confidence bound in MR-UCB,” explains Dr Lee, speaking of their work, which was published in the IEEE Transactions on Neural Networks and Learning Systems on 14 September 2022.
The researchers next derived gap-dependent and independent upper bounds of the cumulative regret. For both the proposed methods, the latter value matches the lower bound under the heavy-tailed noise assumption, thereby achieving minimax optimality.
Further, the new methods require minimal prior information and depend only on the maximum order of the bounded moment of rewards. In contrast, the existing algorithms require the upper bound of this moment a priori–information that may not be accessible in many real-world problems.
Having established their theoretical framework, the researchers tested their methods by performing simulations under Pareto and Fréchet noises. They found that MR-UCB consistently outperformed other exploration methods and was more robust with an increase in the number of actions under heavy-tailed noise.
Further, the duo verified their approach for real-world data using a cryptocurrency dataset, showing that MR-UCB and MR-APE were beneficial–minimax optimal regret bounds and minimal prior knowledge–in tackling heavy-tailed synthetic and real-world stochastic MAB problems.
“Being vulnerable to heavy-tailed noise, the existing MAB algorithms show poor performance in modelling stock data. They fail to predict big hikes or sudden drops in stock prices, causing huge losses. In contrast, MR-APE can be used in autonomous trading systems with stable expected returns through stock investment,” comments Dr Lee, discussing the potential applications of the present work.
“Additionally, it can be applied to personalized recommendation systems since behavioural data shows heavy-tailed noise. With better predictions of individual behaviour, it is possible to provide better recommendations than conventional methods, which can maximize the advertising revenue,” he concludes.
By Adedapo Adesanya
Cloudflare Incorporated, a business providing cloud-based services to various enterprises, said in a note on Friday it is investigating issues with its Dashboard and related Application Programming Interfaces (APIs).
Numerous companies and services, including payments platform like OPay as well as Canva, Coinbase Global Incorporated, Investing.com , Shopify Incorporated, and Zoom Video Communications Incorporated, all appeared to crash, with some seeing “500 internal server error” and “Please check your internet connection and try again”.
The global outage has left many users unable to access these key services as this disruption has not only affected individuals but also businesses relying on these platforms for their operations.
Customers using the Dashboard or Cloudflare APIs are impacted as requests might fail and errors may be displayed, the company said on its status page.
In its latest update, Cloudflare added that “a fix has been implemented,” with the firm monitoring the results.
Users from all over the world have taken to social media platform X (formerly Twitter) to voice their frustrations over the issue.
This is Cloudflare’s second major disruption in nearly a month, following another incident in November that affected services like Spotify and ChatGPT.
At the last outage, Cloudflare’s services were largely restored within three hours, and fully restored after approximately five hours.
By Modupe Gbadeyanka
As part of broader Africa-focused Artificial Intelligence (AI) initiatives, Google has launched the AI Skilling Blueprint for Africa, designed to help governments build a future-proof workforce.
The programme provides governments with a comprehensive, step-by-step guide to formulate national skilling strategies. It focuses on developing three critical cohorts: AI Learners, who will gain foundational AI literacy; AI Implementers, professionals upskilled to integrate AI tools into their work; and AI Innovators, deep technical experts dedicated to building the next generation of AI solutions.
Africa is home to the world’s youngest and fastest-growing population. The continent shows immense potential for AI-driven economic growth.
However, new research highlights a significant challenge: while optimism for AI is exceptionally high, reaching 95 per cent in Nigeria and 76 per cent in South Africa, 55 per cent of firms across the continent report needing AI talent more than financing. Closing this skills gap is key to unlocking Africa’s opportunity.
Google’s Vice President of Government Affairs and Public Policy, Doron Avni, explained that, “The AI Skilling Blueprint provides a clear roadmap for governments to build the workforce of the future.
“By also investing in AI-ready data and expert local organisations and partners, we are helping build the interconnected ecosystem needed for a prosperous, AI-driven future for the continent.”
As part of its broader initiatives, Google also announced $2.25 million to support projects building trustworthy public data sets for AI by the UN Economic Commission for Africa (UNECA), the UN Department of Economic and Social Affairs (UN DESA) and PARIS21.
This contribution will help national statistical offices modernize their infrastructure and empower decision-makers with the reliable data they need to address challenges from food security to economic growth.
“For Africa to drive sustainable development, evidence-based policymaking is indispensable. This requires accessible, reliable, and AI-ready data.
“This effort is a crucial step forward. By building a Regional Data Commons, we can empower African institutions with the data and tools they need to make strategic choices that will drive growth and prosperity,” the Executive Secretary of the UN Economic Commission for Africa, Claver Gatete, said.
Finally, building on its $7.5 million Google.org Skilling Fund commitment, Google announced the first set of expert social impact organizations who will receive funding to execute on projects consistent with its skilling mission, including FATE Foundation and the African Institute for Mathematical Sciences (AIMS), which will embed advanced AI curricula into universities; and JA Africa and CyberSafe Foundation, which will advance crucial work in online safety and digital literacy.
“We are incredibly proud to partner with the African Institute of Management Sciences on the Advanced AI UpSkilling Project, with support from Google.org. This groundbreaking initiative is a direct response to the urgent need for deep AI competencies in Africa, empowering tertiary institutions, lecturers, and students in Nigeria, Ghana, Kenya, and South Africa.
“This strategic support aligns perfectly with FATE Foundation’s mission to foster innovation and sustainable economic growth across the continent, ensuring Africa is fully equipped to lead in the global technological future,” the Executive Director for FATE Foundation, Adenike Adeyemi, stated.
“We live in an age defined by rapid technological change and our mission at JA Africa is to ensure that African youth are not left behind. However, even as we engage our youth in more digital programs and encourage AI literacy, we are fully aware of the harmful effects of unchecked online exposure and, therefore, invest equally in protecting their data, physical safety and mental wellbeing.
“Through this support from Google.org, we will give young people the tools, knowledge, and confidence they need to navigate the digital world safely and responsibly,” the chief executive of Junior Achievement Africa, Simi Nwogugu, remarked.
By Modupe Gbadeyanka
Global technology firm, Zoho, has enhanced its all-in-one business software platform known as Zoho One with improve security, and deeper intelligence across all over 50 applications.
The company improved the user interface, placing context at the centre of the user journey and removes traditional boundaries between applications.
Spaces now organise tools by purpose—such as Personal, Organisation, and Department-specific groups—enabling employees to access what they need without switching between apps. A centralised search bar spans the entire ecosystem, allowing users to find information or trigger workflows instantly.
An enhanced Action Panel provides a full view of upcoming meetings, unread messages, pending tasks, and other key updates, helping employees remain informed regardless of which app they are using.
The updated Dashboard consolidates data from Zoho and third-party apps into one central hub that can be customised using pre-existing or bespoke widgets.
The platform also introduced Vani, a new visual-first collaboration space that supports brainstorming, planning, and creation through diagrams, whiteboards, mind maps, and integrated video calling.
A central integrations panel enables administrators to monitor and configure all connections. Foundational integrations bring application-specific portals—Zoho or third-party—into a single unified portal. Practical tasks such as domain verification and authentication can now be configured more easily.
The new Smart Offboarding feature introduces outcome-based integrations, allowing organisations to transfer department ownership, manage employee device data, and determine data access rights within a single workflow, ensuring smooth transitions.
Also, Zia, Zoho’s AI assistant, is now accessible throughout Zoho One, providing unified intelligence that supports decision-making and improves productivity. Zia can aggregate and contextualise information from various platforms, including third-party systems such as Google Workspace, and present it as clear, actionable insight.
Zia Hubs, the platform’s intelligent content management system, now has a dedicated space where contracts, meeting recordings, and other important assets are automatically organised. Through Zia Search, employees can quickly surface relevant information without navigating multiple locations.
In addition, Ask Zia, available from the bottom toolbar, enables prompt-based searches across Zoho One, providing quick visibility into schedules, tasks, recent interactions, and other key details.
Commenting on the changes, the Country Head for Zoho Nigeria, Mr Kehinde Ogundare, said, “The Zoho One update reflects how work has evolved from using individual applications to operating within a unified platform.
“Zoho One customers are not simply licensing apps; they are choosing a solution that allows Zoho to handle the technology while they focus on productivity. The enhancements announced today deliver a cohesive experience built on unified integrations, context, and data.”
-
Feature/OPED6 years agoDavos was Different this year
-
Travel/Tourism9 years ago
Lagos Seals Western Lodge Hotel In Ikorodu
-
Showbiz3 years agoEstranged Lover Releases Videos of Empress Njamah Bathing
-
Banking7 years agoSort Codes of GTBank Branches in Nigeria
-
Economy2 years agoSubsidy Removal: CNG at N130 Per Litre Cheaper Than Petrol—IPMAN
-
Banking3 years agoFirst Bank Announces Planned Downtime
-
Banking3 years agoSort Codes of UBA Branches in Nigeria
-
Sports3 years agoHighest Paid Nigerian Footballer – How Much Do Nigerian Footballers Earn