The Role of GANs, VAEs, and Autoregressive Models in

Neurological and Psychological Research: A Comprehensive

Review

Mona Yassen1,*

1 Computer Science and Intelligent Systems Research Center, Blacksburg 24060, Virginia, USA

Received: January 17, 2026 Revised: February 15, 2026 Accepted: April 15, 2026 ⋆ Corresponding author

ABSTRACT

Generative models, including GANs, VAEs, and autoregressive models, have drastically improved neurological

and psychological analyses. They allow for the formation of complex data representations that imitate cognitive

processes in a human brain and thus help to understand the workings of the brain and mental diseases. Thus,

GANs trained with adversarial mechanisms can synthesize nearly photorealistic fake samples, which can mimic

neurological disorders or assess the effectiveness of treatment. VAEs are known to give a formidably well-founded

method of learning the hidden representations of psychological states, therefore allowing researchers to study the

potential causes of mental health problems. Autoregressive models, in contrast, are most applicable in time series

data, which is highly important when the neurological signal or behavior under investigation needs to be studied

over time. This broad survey discusses the assets and liabilities of these generative approaches, emphasizing their

usability in simulating elaborate psychological processes and deploying evidence-based observations to understand

the assessment and therapy of mental disorders. Therefore, this work aims to reveal the significant connections

between these methodologies and their further potential for investigating human cognition and behavior through the

coordinated usage of highly effective computational methods.

Keywords: Generative Models GANs Autoregressive Models Neuroscience Mental Health Disorders

1. INTRODUCTION

In recent years, artificial intelligence has undergone significant

development, leading to groundbreaking innovations in

fields such as neuroscience and psychology [1, 2, 3]. Among

these advancements, generative models, including Generative

Adversarial Networks (GANs), Variational Autoencoders

(VAEs), and autoregressive models, are considered highly influential

[4, 5]. These methods enable researchers to analyze

intricate patterns that characterize complex datasets and have

the potential to transform current understanding of the human

brain.

The human brain is characterized by a high degree of complexity,

which raises numerous scientific questions. Generative

models can simulate brain-related processes and different

psychological states, thereby supporting a deeper understanding

of psychiatric disorders and cognitive mechanisms. These

tools can help researchers reconstruct and simulate brain activity

and identify possible treatments for neurological disorders

[6].

Neural networks that use adversarial training, known as Generative

Adversarial Networks, have become particularly important

in neuroscience. By learning from high-quality data,

GANs can generate samples that resemble neurological con-