| viXra.org > Artificial Intelligence > 2412 |
 |
 |
| viXra.org > Artificial Intelligence > 2412 |
|
|
[9] viXra:2412.0166 [pdf] replaced on 2025-01-05 09:05:25
Authors: Satish Gajawada
Comments: 2 Pages.
Particle Swarm Optimization (PSO) is a popular and widely used optimization algorithm for solving complex problems. It is known for its simplicity and ease of implementation. Artificial Birds move in search space to find optimal solution. Although many PSO algorithms were proposed in literature the concepts like happiness and health are not yet explored in PSO algorithms. This article is based on this research gap. Happiness and Health Particle Swarm Optimization (HaHePSO) algorithm is created by incorporating the Happiness and Health concepts into Particle Swarm Optimization algorithm. Each particle in HaHePSO algorithm is associated with happiness and health variables. The movement of Artificial Birds in PSO algorithm is based on fitness values. In HaHePSO algorithm the movement of Artifical Birds is dependent on happiness, health and fitness values. In PSO algorithm Artificial Birds move in the direction of local best and global best of fitness values. This idea is extended in HaHePSO algorithm where Artificial Birds move in the direction of local best and global best of happiness, health and fitness values. The HaHePSO algorithm proposed in this article takes more space and requires extra computation compared to PSO algorithm. This is due to the fact that each particle now has happiness and health variables associated with it and movement in search space is guided by the fitness, happiness and health values.
Category: Artificial Intelligence
[8] viXra:2412.0149 [pdf] submitted on 2024-12-24 01:32:10
Authors: Stephane H. Maes
Comments: 14 Pages. All related details of the projects (and updates) can be found and followed at https://shmaes.wordpress.com/
In October and November 2024, using popular LLMs like OpenAI ChatGPT (4 and below), Azure OpenAI and its Copilot instantiations, Google Gemini and GenAI LLM tuned for scientific papers like Zendy, asking a question and references produces with every LLM fake references, well onstructed, but with different titles or authors than the web or journal reference actually associated to the citation, or sometimes totally invented. Prompting to ensure that the reference exists and is correct may help for some, but in general it does not. Others have reported similar issues when using these LLM/GenAI services to produce legal briefs, and other legal documents.This paper suggest simple ways to address this, instead of trying to just improve the LLMs and hope hallucinations will be reduced; they won’t, no matter what, they are inherent to LLM. It is very surprising and mindboggling that LLM providers have not been implementing these kind of solutions: just check if the references exist, are correctly cited, and relevant to the paper/context. We also expand the approach with our MultiAI approach to improve on the previous approach, or address other hallucinations; actually eliminating in our tests.
Category: Artificial Intelligence
[7] viXra:2412.0140 [pdf] submitted on 2024-12-22 14:24:26
Authors: Danil Kutny
Comments: 6 Pages.
This paper introduces a modification to standard GPT-like models by incorporating character-level encoding. The model uses an LSTM to process individual characters within tokens, which are then embedded into the original token embedding space. This allows the model to maintain token-level processing while adding character-level information to each token. Trained on the BookCorpus dataset, the model was evaluated on tasks requiring character-level manipulation, such as counting letters and reversing words. Surprisingly, the modified model performed similarly to the baseline GPT model, with no significant improvements, suggesting that GPT-like models may inherently learn character-level representations from tokenized inputs.
Category: Artificial Intelligence
[6] viXra:2412.0129 [pdf] submitted on 2024-12-22 03:12:27
Authors: Mohammadjavad Maheronnaghsh, Taha Akbari Alvanagh
Comments: 20 Pages. This article will be published in ECCV by Springer.
The persistence of spurious features in machine learning models remains a significant challenge. To address this issue, we identify several future directions that require attention. Firstly, we highlight the need for a new dataset that allows researchers to control the types and levels of spurious features, as this resource is currently lacking. Secondly, we emphasize the importance of addressing spurious features in natural language processing, where more attention is needed compared to vision-related tasks. We also stress the need for addressing spurious correlations at the core algorithmic level, rather than relying on complex, task specific solutions that may not generalize well. Finally, we advocate for the development of weakly-supervised or unsupervised methods that reduce reliance on group labels, making the approaches more widely applicable. Our review aims to provide a comprehensive overview of existing work and guide future research in creating more robust machine learning models.
Category: Artificial Intelligence
[5] viXra:2412.0114 [pdf] submitted on 2024-12-19 15:45:16
Authors: Satish Gajawada
Comments: 2 Pages.
The idea is to incorporate the concept of money into Particle Swarm Optimization (PSO) algorithm to create a new PSO algorithm titled "Money Particle Swarm Optimization (MyPSO)" algorithm.
Category: Artificial Intelligence
[4] viXra:2412.0057 [pdf] submitted on 2024-12-09 21:29:21
Authors: John Tian
Comments: 9 Pages. Distributed under the CC BY license
Continual learning (CL) enables machine learning models to learn tasks sequentially while maintaining performance on previously learned tasks. This capability is crucial for developing intelligent systems that adapt to evolving conditions across domains like robotics, recommendation systems, and autonomous vehicles. However, neural networks typically suffer from catastrophic forgetting, where learning new tasks disrupts performance on older ones, often necessitating costly retraining from scratch.We present GKD-ER (Gradient-space Knowledge Distillation with Episodic Replay), a framework that effectively reduces catastrophic forgetting by combining three complementary techniques:
Gradient Projection (GP): Removes gradient components that would harm older tasks, ensuring parameter updates for new tasks remain orthogonal to previously learned knowledge.
Knowledge Distillation (KD): Maintains functional consistency by aligning the current model's outputs with those from a saved reference model on old data.
Episodic Replay (ER): Periodically revisits representative samples from past tasks stored in a memory buffer, reinforcing old decision boundaries and providing stable checkpoints.
Under standard conditions and representative replay assumptions, we theoretically demonstrate that GKD-ER achieves bounded forgetting. Our empirical evaluation on established benchmarks like Permuted MNIST and Split MNIST shows that GKD-ER surpasses strong baselines (Naive, EWC, SI, and ER alone) with higher final accuracies, significantly reduced forgetting, and stable class-level decision boundaries across tasks.By integrating constraints at the gradient, functional, and empirical levels, GKD-ER strikes an effective balance between stability and plasticity. This work advances the development of systems capable of continuous learning while preserving past expertise—a key step toward truly adaptive, lifelong learning agents.
Category: Artificial Intelligence
[3] viXra:2412.0049 [pdf] submitted on 2024-12-09 20:27:27
Authors: Akira Pyinya
Comments: 10 Pages.
This article briefly describes a new definition of intelligence: Doing the same thing in new situations as the examples of the right thing to do, by making predictions based on these examples. In other words, intelligence makes decisions by stare decisis with Solomonoff induction, not by pursuing a final goal or optimizing a utility function. This general theory of intelligence is inspired by Assembly theory, the Copycat model, and the Active inference approach, and is formalized using Algorithmic information theory.
Category: Artificial Intelligence
[2] viXra:2412.0021 [pdf] submitted on 2024-12-05 07:16:25
Authors: Eugene Rulko
Comments: 9 Pages.
Current level of development in the sphere of deep learning allows replacing existing domain-specific algorithms for military simulation with approximating neural networks. Hyperparameter search allows finding network’s architecture, appropriate for a task. This work describes that process for the task of predicting area of optical visibility, taking a fragment of a digital map as input and proposes ancillary architectural solutions for stitching building blocks together, assuring their conformation for performing search among their possible combinations within the architectural space. The final proposed result is a channel-wise attention U-Net with an encoder, based on ResNet50 backbone.
Category: Artificial Intelligence
[1] viXra:2412.0019 [pdf] submitted on 2024-12-05 08:05:02
Authors: Clarence Antipona, Romeo Magsino, Raymund Dioses, Khatalyn Mata
Comments: 9 Pages.
This study is focused on enhancing the Haar Cascade Algorithm to decrease the false positive and false negative rate in face matching and face detection to increase the accuracy rate even under challenging conditions. The face recognition library was implemented with Haar Cascade Algorithm in which the 128-dimensional vectors representing the unique features of a face are encoded. A subprocess was applied where the grayscale image from Haar Cascade was converted to RGB to improve the face encoding. Logical process and face filtering are also used to decrease non-face detection. The Enhanced Haar Cascade Algorithm produced a 98.39% accuracy rate (21.39% increase), 63.59% precision rate, 98.30% recall rate, and 72.23% in F1 Score. In comparison, the Haar Cascade Algorithm achieved a 46.70% to 77.00% accuracy rate, 44.15% precision rate, 98.61% recall rate, and 47.01% in F1 Score. Both algorithms used the Confusion Matrix Test with 301,950 comparisons using the same dataset of 550 images. The 98.39% accuracy rate shows a significant decrease in false positive and false negative rates in facial recognition. Face matching and face detection are more accurate in images with complex backgrounds, lighting variations, and occlusions, or even those with similar attributes.
Category: Artificial Intelligence
| Third-Party Links: |
|