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 Modupe Gbadeyanka
As part of its commitment to celebrating and rewarding creativity, Facebook has updated its guidance, with clear definitions of what counts as original and unoriginal content.
In a message on Monday, the social media platform said it was offering content creators new tools to report impersonation.
Launched last year, the content protection tool is expanding beyond detecting reel matches across Meta platforms to now also flag potential impersonation.
Creators can take action on content theft and easily submit impersonation reports all in one place.
Facebook, in the statement received by Business Post, said creators can check for access to content protection in their professional dashboard or apply for access here.
The platform also disclosed that in 2025, it removed over 20 million accounts impersonating large content creators, and impersonation reports related to large content creators dropped by 33 per cent.
Further, Facebook is deprioritising unoriginal content by making sure they do not perform well on its platform.
It noted that content that is duplicated from other sources or makes low-value changes to someone else’s content may see significantly reduced reach, and accounts that primarily post unoriginal content may lose eligibility for recommendations and monetisation.
It was emphasised that “these changes provide creators who post original content with greater reach and monetisation opportunities, provide stronger protections for their work, and reduce the reach of unoriginal content.”
By Dipo Olowookere
A new high-density appliance that enables enterprises to scale cloud-managed physical security without forcing cloud-only storage or infrastructure replacement has been launched by a global leader in enterprise physical security software, Genetec.
The product, Cloudlink 2210, was designed for complex, enterprise-scale deployments and supports multiple workloads, including video management, access control, and intrusion detection, in a single appliance. By consolidating these workloads into one appliance, it reduces system sprawl, simplifies management in large-scale environments, and lowers operational overhead.
Unlike solutions that separate workloads across multiple proprietary systems, Genetec Cloudlink 2210 is built on an open architecture that supports a wide range of third-party devices, including cameras, access control systems, and intrusion panels. This enables organisations to modernise at scale within a unified, cloud-managed model designed to preserve architectural flexibility, while securely integrating existing hardware, maintaining business continuity, and reducing migration risks.
The company disclosed that Cloudlink 2210 also supports hundreds of connected devices per appliance and provides up to 240 TB of local storage per unit, making it well-suited for deployments with high device density and long retention policies. The Cloudlink 2210 is ideal for enterprise environments where uptime and local retention requirements are operational priorities because its design minimises dependence on cloud storage, helping organisations control long-term storage costs while maintaining the performance and availability required in enterprise environments.
The new product also incorporates hardware-level resiliency to support strict uptime and retention requirements. RAID-protected storage and redundant system components help ensure data protection and OS availability. Security workloads continue operating locally, independent of cloud connectivity, allowing deployments to maintain continuity even during network disruptions. Dual network interfaces provide redundancy and support network isolation to strengthen cybersecurity.
It scales by adding units as requirements grow, enabling organisations to increase device counts and storage capacity without redesigning their infrastructure. Centralised cloud management maintains visibility and control across deployments.
Genetec Cloudlink 2210 is part of the broader Genetec approach to deployment flexibility. The cloud-managed appliance portfolio enables organisations to operate on premises, in the cloud, or across hybrid environments based on their operational and regulatory requirements. By combining high-performance local processing and storage with centralised cloud operations and management, Cloudlink 2210 supports scalable, cloud-managed deployments without compromising control or performance.
The Product Director for Unified Solutions at Genetec Incorporated, Mr Christian Chenard Lemire, said, “Enterprises don’t want to choose between innovation and operational certainty.
“With Cloudlink 2210, we’re redefining what cloud-managed physical security looks like at scale by giving organisations the freedom to modernise on their own terms, control long-term costs, and maintain the resiliency and continuity their most critical environments demand.”
By Modupe Gbadeyanka
An additional $200,000 will be invested in Artificial Intelligence (AI) media literacy initiatives across Sub-Saharan Africa, TikTok announced during its third annual Sub-Saharan Africa Safer Internet Summit in Nairobi, Kenya.
The platform hosted government officials, regulators, online safety partners and industry leaders for the event, reinforcing its commitment to collaborative approaches to online safety.
The funds will be provided in ad credits to help support local organisations in the region to expand AI media literacy.
This investment builds on the company’s initial $2 million AI Literacy Fund, launched in November 2025, which awarded 20 global non-profits to create content that boosts public understanding of AI.
In Sub-Saharan Africa, TikTok initially supported three organisations to advance digital literacy and combat misinformation.
“With the rapid advancement of AI, we are committed to educating our community online, so they feel empowered to have responsible experiences with AI, whether that’s as viewers or creators.
“We are partnering with trusted local organisations that communities already know and rely on, because their expertise and deep local connections are essential to making AI literacy programs truly impactful,” the Global Head of Partnerships, Elections and Market Integrity at TikTok, Mr Valiant Richey, stated.
Earlier, the Head of Government Relations and Public Policy for Sub-Saharan Africa at TikTok, Ms Tokunbo Ibrahim, said, “As we host the 3rd Annual Safer Internet Summit here in Kenya, our mission is clear: to share learnings, insights, tackle common challenges and collaboratively advance actionable solutions that protect citizens online.
“By bringing together a diverse coalition of policymakers, tech innovators, and creators, we are ensuring that the conversations we have at this Summit are all-inclusive and lead to a more resilient digital landscape.”
The summit featured expert panels and discussions on critical topics, including TikTok’s Trust and Safety efforts, protecting young people online, and policy frameworks for responsible AI governance.
A key highlight of the event was showcasing how TikTok uses AI to transform how people share their creativity and discover new passions, while ensuring the community remains safe through transparent and responsible AI practices.
The platform also shared more about how recent advancements in AI are helping the platform moderate content faster and more consistently at scale, by improving automated moderation and empowering human teams with better moderation tools.
With over 100 million pieces of content uploaded daily to TikTok, these advances, which work alongside human moderation teams, are helping get violative content down faster, reducing the likelihood of the community seeing it.
According to the latest Community Guidelines Enforcement Q3 2025, TikTok removed over 14 million videos across Sub-Saharan Africa, with 96.7 per cent detected and removed proactively using automated technology, underscoring TikTok’s commitment to proactive moderation and swift action.
-
Feature/OPED6 years agoDavos was Different this year
-
Travel/Tourism10 years ago
Lagos Seals Western Lodge Hotel In Ikorodu
-
Showbiz3 years agoEstranged Lover Releases Videos of Empress Njamah Bathing
-
Banking8 years agoSort Codes of GTBank Branches in Nigeria
-
Economy3 years agoSubsidy Removal: CNG at N130 Per Litre Cheaper Than Petrol—IPMAN
-
Banking3 years agoSort Codes of UBA Branches in Nigeria
-
Banking3 years agoFirst Bank Announces Planned Downtime
-
Sports3 years agoHighest Paid Nigerian Footballer – How Much Do Nigerian Footballers Earn