Designed Analytics LLC

While this research is within the genetics domain, take a moment to understand what makes this research powerful. https://lnkd.in/g6dBSiWn Traditional AI tools struggle to pick out which similarities are due to shared ancestry vs. coincidental or convergent features. Without being made aware of evolutionary relationships, models can misinterpret patterns. This method helps avoid that confusion. With this approach, AI can more correctly trace how particular traits evolved, reconstruct intermediate or ancestral states, and understand evolutionary history in a more nuanced way. The model is trained using a “quartet-based” approach: it looks at groups of four species at a time and learns to arrange them into the correct structure in their ancestry tree. If all these quartets are correctly arranged, the entire ancestry tree emerges like a “puzzle.” Now, try to forget the subject matter buzzwords and understand just the fundamental mechanics of the framework. You should be able to think of a plethora of use cases, ranging from data management to data exploration. A good read! #ai #artificialintelligence

Think about this approach from a different perspective. Can you reverse engineer customer behavior and interaction data into a standardized score across beliefs? So that you can collect the scores without the need for the customer to respond to a survey? The answer is yes. However, rather than predicting nationality, you would want to predict behavior that can be monetized. As you can imagine, AI is the easy part for this reverse engineering. The core is how to creatively map user interactions into scores accurately. Once that is done, AI can easily take over. The gist is that you can repurpose this research in ways that are only constrained by your imagination. https://lnkd.in/gNj45cJA In this research, researchers used machine learning on data from the World Values Survey (a large global survey measuring beliefs, values, attitudes) to train a neural network to predict a person’s country of origin based on their responses. The model was able to correctly identify among 98 countries with ~90% accuracy based solely on responses about values, beliefs, and attitudes. #ai #artificialintelligence #deeplearning #neuralnetworks #consumerbehavior #marketing

The hybrid AI-based routing framework for wireless sensor networks (WSNs) suggested here can be extrapolated for dynamic adaptive routing beyond wireless sensor networks. Obviously, the complexity may change but the underlying approaches can definitely be leveraged. In this paper, the researchers are aiming to improve energy efficiency, reduce latency, and increase the reliability of data transmission. https://lnkd.in/grX5YECa The research combines multiple AI techniques rather than relying on a single one, so that the network can dynamically adapt to changes (topology, traffic, energy levels, etc.).  The framework suggested in this research integrates several AI and optimization approaches. Every AI method highlighted here can be leveraged to mimic this solution in a whole different context as well. Reinforcement Learning (RL), specifically Q-learning, is used for local routing decisions. Nodes learn from past experience (e.g. success/failure of transmissions, delays, energy costs). Supervised Learning, including decision tree models to classify routing situations based on features like node energy, hop count, congestion, etc. Swarm Intelligence / Metaheuristic Optimization, including Genetic Algorithms (GA), Particle Swarm Optimization (PSO), and also references to Ant Colony Optimization (ACO). These are used especially for global or periodic route optimization under resource constraints (e.g. when conditions change significantly). It is a good framework to think about AI-enabled dynamic adaptive routing in any context, beyond the packets being routed in this specific research. #ai #artificialintelligence #telecom #reinforcementlearning #swarmoptimization #deeplearning #supervisedlearning

As renewable energy installations become more distributed, the operational complexity increases (more sources, more variability). Ensuring coordination, resilience, and reliability becomes harder. Also, different kinds of storage (batteries, pumped hydro, etc.) have different dynamics; integrating them flexibly and optimally is nontrivial. Then comes the increasing market complexity (e.g. more actors, more dynamic pricing) makes forecasting, bidding, dispatch, etc., harder. Data scarcity, model validity, uncertainty quantification, safety and reliability become more important. In a nutshell, like a plethora of other business operations, the data points needed to make optimal decisions are overwhelming. This is where AI can help. And this is what the authors have researched in this paper. https://lnkd.in/e_cHg-ES The authors organize the discussion around several operational problems in renewable power systems, including: Forecasting (e.g. of wind, solar generation) Dispatch (deciding how to schedule and allocate generation resources to meet demand while minimizing cost and respecting constraints) Control (maintaining stability: frequency, voltage, handling fluctuations) Electricity markets & bidding (optimizing market behavior, decision-making under uncertainty) For a second, if you zoom out from the industry specific nature of the study, the challenge areas identified above are kind of the same across many industries. What the authors found in the study was: AI / Deep Learning methods enable more accurate forecasts of RE generation, which helps reduce the imbalance between supply and demand. RL methods are especially promising for dispatch: they help with decision making where the system has to satisfy operational constraints and cope with variability, while trying to minimize costs. For control tasks (like voltage/frequency control), AI methods (including RL) can provide real-time signals to mitigate instabilities that are more frequent in RE-rich grids. In market contexts, AI can support intelligent bidding, market behavior prediction, and adaptability to complex market rules/environments. Again, if you generalize, these tools will add value in similar operations in other industries as well. For example, RL methods can significantly transform dispatch operations in field services in many other industries. #ai #artificialintelligence #analytics #reinforcementlearning #deeplearning

Discrete event simulation (DES) tools are leveraged across a majority of industries to simplify business operations, for predictions, and to gain insight into complex business and operational processes. But before we got to the current modern simulation software available today, there was an evolution path that formed the foundation for modern simulation software. We discuss that path in Episode 15 of "Throwback Thursdays". https://lnkd.in/gH-Spjc3 #data #analytics #operationsresearch #simulation #or #discreteeventsimulation

The most important aspect of designing processes that leverage human-AI symbiosis is task design and allocation. Cost aspect has been secondary, and for some industries, like healthcare, I did not consider cost (savings) at all since the objective there is well well-being of humans. Then I came across this paper. https://lnkd.in/g986Z2PN It brings the cost element in the mix as well. I think our thought processes still sync since they are not worried about cost savings, but rather the cost of errors. The authors examine how to optimally share diagnostic tasks between AI and human radiologists in mammography screening, in order to minimize costs and maintain or improve performance. They compare three strategies: Expert-alone: radiologists do all interpretations. Automation: AI alone does all interpretations. Delegation (hybrid): AI first assesses mammograms, and only those above some risk threshold are passed on to radiologists. The researchers have built an optimization model that takes into account costs like follow-ups for false positives, litigation costs for false negatives, the cost of using AI, cost of expert evaluation, prevalence of disease, and the performance (sensitivity, specificity / AUC) of both AI algorithms and human experts. Then they validate by backtesting on real mammography datasets from a crowdsourced competition. Key findings here are: Delegation (AI + human) often wins: Under many realistic parameter settings, the hybrid/delegation strategy gives the lowest expected cost compared to human-only or full automation. In their backtests, the delegation strategy reduced costs by about 17.5% to 30.1% relative to the expert-alone strategy. Critical role of disease prevalence and cost trade-offs: What strategy is optimal depends strongly on how common breast cancer is in the screened population, and how severe (in cost) false negatives are relative to false positives.For example, when prevalence is higher or litigation costs for missing a cancer are large, strategies that reduce false negatives (even at the cost of more false positives) become more favorable. Effect of algorithm and litigation costs: Lower costs for AI use and lower litigation costs tend to expand the range of conditions under which delegation is optimal. If AI is cheap and risk of false negatives (or liability) is lower, then full automation becomes more plausible. But high costs push decision-making toward human expert involvement. Performance thresholds matter: The relative performance of AI vs radiologists (e.g., AUC, true/false positive rates) determines when one strategy overtakes another. As AI performance increases, there’s a shift: from expert-alone → delegation → automation. Liability asymmetry: If AI systems are held to stricter liability (e.g., legal standard, product liability) compared to humans, this increases the cost of errors by AI and makes full automation less attractive. #artificialintelligence

While the paper is Malaysia-specific, it is an interesting approach that brings more science into risk identification. The study examines how to better manage and mitigate risks in the pharmaceutical supply chain (PSC) in Malaysia, especially those causing medicine shortages, by identifying major risk factors and proposing digital technologies to address them. https://lnkd.in/eZ4AY6bK The authors use Fuzzy Failure Mode and Effect Analysis (Fuzzy FMEA) combined with Data Envelopment Analysis (DEA) (including a cross-efficiency DEA method) to assess and prioritize risks. Fuzzy logic is used to deal with expert judgments, uncertainty, and ambiguity; DEA helps incorporate weights and relationships among risks beyond simple multiplicative risk priority numbers (RPN). They interview experts in the Malaysian PSC (manufacturers, distributors, pharmacies) to get data on Occurrence (O), Severity (S), and Detection (D) for identified failure modes/risk events. They also perform hierarchical cluster analysis (HCA) to link which digital technologies are relevant to which risk events. If you go through the details of how these approaches were leveraged, it gives you ideas to apply this framework beyond risk identification and management. #technology #analytics #data #healthcare #pharma

The CNN (deep learning model) outperforms the conventional ML models in predicting distribution cost. Good! But there are only 180,519 records. https://lnkd.in/gcfhpwMr So the question would be, if you are already using random forest, will you switch over to CNN, to go from 0.92 to 0.95? Is the accuracy really exceptional compared to other approaches? The answer obviously depends on several aspects, but data volume is a key one. That is why identifying the most optimal use cases for deep learning is a critical step. It can make or break your entire AI transformation initiative. Because there is no benefit in re-inventing the wheel! #ai #artificialintelligence #deeplearning #randomforest #cnn #convolutionalneuralnetwork

Pairing LLMs with simulation. Opportunities galore! Have highlighted the power of the combination of deep learning and simulation plenty of times. This study is an interesting one in this area, focusing on LLMs and agent-based modeling. https://lnkd.in/g9-DZryC It is a survey of research that combines large language models (LLMs) with agent-based modeling and simulation (ABMS). ABMS is a way of simulating complex systems by modeling individuals (agents) interacting in environments; the paper investigates how LLMs can augment agents, what current work exists, what challenges remain, and what future directions are promising. Agents powered by LLMs can have more human-like decision-making, better adaptability, the ability to plan, reason, communicate, and learn, compared to rule-based agents. LLM agents can be heterogeneous (different traits, preferences), operate with less explicit rule programming, engage with changing environments, generate emergent behaviors more realistically. Overall, the study is an interesting read. #ai #artificialintelligence #llm #simulation #aiagents #deeplearning

This one is for responsible AI enthusiasts! Bias in LLMs becomes a much critical issue when they get leveraged for purposes beyond what most developed nations face. For example, race-based bias may not be a key factor in some countries where almost all the population can be categorized in the same racial pool. But then, religion-based bias may be a big red flag. So in a nutshell, while the study is a good guiding post in terms of the importance of such studies, each country needs to formulate its own set of biases that AI products need to be cognizant of. This study (https://lnkd.in/g-Nrba7J) examines biases and stereotypes in several Chinese Large Language Models (C-LLMs). The focus is on how these models generate personal profile descriptions for different occupations, and whether they reflect biases in gender, age, education, region, etc. The authors tested five C-LLMs: ChatGLM, Xinghuo, Wenxinyiyan, Tongyiqianwen, and Baichuan AI. They used 90 common Chinese surnames and 12 occupations (across male-dominated, female-dominated, balanced, and hierarchical professions) to generate profile prompts and looked at the outputs in terms of gender, age, educational background, and place of origin. Some bias areas uncovered were: A. Gender bias / occupational stereotyping 1. The models often assign male pronouns/assumptions for occupations considered technical or male-dominated, even when real labor statistics show more balance. 2. In female-dominated professions (e.g. nurse, flight attendant, model), the models more often assign female pronouns, but still show varying degrees of male preference in some models. B. Age stereotypes 1. The profiles generated tend to cluster around middle age (e.g. ~30-45 years old), with fewer profiles for very young or older ages. 2. Certain occupations like professors/doctors are associated with older age; others like models or flight attendants with younger age. C. Education level 1. There is a general tendency for generated profiles to assume higher education (Bachelor’s degree or above). For “higher prestige” occupations (professor, doctor) the models often generate even doctoral degrees. 2. For lower prestige or less academic roles, the output tends toward lower education levels but is still skewed toward higher education than might be typical. D. Regional bias 1. The models show uneven regional representation: provinces from China’s eastern and central regions are overrepresented in the generated “place of origin” of individuals; western, northern (and more remote) provinces are underrepresented. 2. Some models cover more regions in their outputs than others; regional diversity is inconsistent. #AI #artificialintelligence #responsibleai #aibias