The Chinese Journal of Artificial Intelligence - Current Issue

Artificial intelligence postulates that computers will eventually supervise performing tasks through various pattern recognition with less or without human interventions and assistance. It appears to mimic human cognitive functions. Resembling the human brain, it receives various forms of raw data that are stored, aligned, surveyed, interpreted, analyzed, and converted to single processed data, making it easy to conclude and understand. Recently, in the digital world, machine learning, deep learning, neural network and AI applications are expanding widely, where humans have expertise.

A detailed literature survey was performed through an online database, such as ScienceDirect, Google Scholar, Scopus, Cochrane, and PubMed. The search keywords were Machine Learning OR Deep Learning OR Neural Networks OR Applications OR Pharmaceutical Innovations OR Technology OR Artificial Intelligence AND Pharmaceutical Sectors OR Clinical Pharmacology OR Healthcare OR Medical OR Pharmacovigilance OR Clinical Trials OR Regulatory OR Challenges. The literature search was limited to studies published in English.

It was found that there is an immense growth of artificial intelligence in the sector of the pharmaceutical industry applied in drug discovery and drug development, clinical trials, and the pharmacovigilance sector. It has several clinical applications of AI as a tool in health care and biomedical research besides clinical practice. It also shows several challenges faced and methods to overcome them.

AI has great potential and future as a valuable tool in the healthcare and pharmaceutical industry by applying a scientific approach and averting real-life challenges.

The purpose of this study is to determine under what conditions, such as noise and malfunction, successful consensus achievement in swarm robotics is possible.

Swarm robots can be used to solve exploration problems, such as the best-of-n problem. Consensus achievement plays a crucial role as the swarm must collectively agree on a solution. This task can be even more challenging considering noise and malfunctioning or rogue agents.

This study aims to determine how robust the consensus achievement algorithm is against noise and rogue agents, considering the effect of adding memory to the agents and further parameter tuning.

We implement a baseline based on the democratic honeybees algorithm and investigate the performance and robustness of the consensus achievement during a number of computational experiments. In particular, the number of agents in the swarm, the number of iterations, the number of positions an agent can visit per iteration, the number of neighbors an agent shares its best option with, and the majority threshold defining the majority based on a fraction of agents in the swarm, and the minimum number of iterations to achieve consensus are investigated regarding their impact.

For better performance, memory has been implemented so that each agent remembers and retains their previous highest quality score if no one better has been found in the current exploration phase. We show that the algorithm is viable and offers robustness in the considered scenarios when memory is added. In particular, we establish a baseline for the democratic honeybees algorithm and ascertain adequate parameter values to ensure the algorithm's best performance. The algorithm is sufficiently robust against noise, and to an extent, against rogue agents. Furthermore, parameter tuning also proved to help the swarm explore very large search spaces.

The consensus algorithm appears sufficiently effective under adverse conditions such as noise and rogue agents, especially when countermeasures are considered.

Further scenarios such as specific communication topologies could be investigated in future research.

Examination Timetabling Problem that tries to find an optimal examination schedule for schools, colleges, and universities, is a well-known NP-hard problem. This paper presents a Genetic Algorithm variant approach to solve a specific examination timetabling problem common in Japanese colleges and universities.

The proposed algorithm uses a direct chromosome representation Genetic Algorithm and implements constraint-based initialization and constraint-based crossover operations to satisfy the hard and soft constraints. An island model with varying crossover and mutation probabilities and an improvement approach called pre-training are applied to the algorithm to further improve the result quality.

The proposed model is tested on synthetic as well as real datasets obtained from Sophia University, Japan and shows acceptable results. The algorithm was fine-tuned with different penalty point combinations and improvement combinations.

The comparison results support the idea that the initial population pre-training and the island model are effective approaches to improve the result quality of the proposed model. Although the current island model used only four islands, incorporating a greater number of islands, and some other diversity maintenance approaches such as memetic structures are expected to further improve the diversity and the result quality of the proposed algorithm on large scale problems.

In computer vision applications, gait-based age estimation across several cameras is critical, especially when following the same person from various viewpoints.

Gait-based age recognition is a very challenging task as it involves multiple hurdles, such as a change in the viewpoint of the person. The proposed system handles this problem by performing a sequence of tasks, such as GEI formation from silhouette, applying DCT on GEI and extracting the features and finally using MLP for age estimation. The proposed system proves its effectiveness by comparing the performance with state-of-the-art methods, conventional methods and deep learning-based methods. The performance of the system is estimated on OU-MVLP and OULP-Age datasets. The experimental results show the robustness of the system against viewing angle variations.

This study aimed to implement the system, which adopts a lightweight approach for gait-based age estimation.

The proposed system uses a combination of the discrete cosine transform (DCT) and multi-layer perceptron (MLP) on gait energy image (GEI) to perform age estimation.

The performance of the system is extensively evaluated on the OU-MVLP and OULP-Age datasets.

The proposed system attains the best mean absolute error (MAE) of 5.05 (in years) for the OU-MVLP dataset and 5.65 for the OULP dataset.