Comparative Analysis of Stochastic Gradient Descent Optimization and Adaptive Moment Estimation in Emotion Classification from Audio Using Convolutional Neural Network
Bibliographic record
Abstract
<ns2:p>Emotion is a fundamental aspect of human life that profoundly shapes behavior, social interactions, and decision-making processes. The ability to effectively communicate and foster mutual understanding between individuals relies heavily on accurately recognizing and expressing emotions. Among various channels of emotional expression, sound stands out as a powerful and direct medium that reflects and conveys human emotional states. This makes audio-based emotion recognition a critical and rapidly evolving field of study. With the rapid advancements in information technology and artificial intelligence, research focused on recognizing emotions through sound signals has gained significant momentum. Machine learning algorithms, particularly deep learning models like neural networks, have demonstrated remarkable capabilities in identifying and classifying emotions expressed through multiple modalities such as text, images, videos, and especially audio signals. Within the family of neural networks, Convolutional Neural Networks (CNNs) have been especially effective for audio emotion classification, due to their strength in extracting hierarchical and spatial features directly from raw input data. This study specifically investigates the comparative effectiveness of two popular optimization algorithms—Stochastic Gradient Descent (SGD) and Adaptive Moment Estimation (Adam)—in training CNN models for emotion classification from audio recordings. Utilizing the Ryerson Audio-Visual Database of Emotional Speech and Song (RAVDESS) dataset, experimental results indicate that CNNs trained with the SGD optimizer achieve an overall accuracy of 53%, surpassing the 48% accuracy achieved by Adam. These results underscore the potential advantages of SGD in fine-tuning deep learning models for audio-based emotion recognition. Consequently, researchers and practitioners are encouraged to consider SGD optimization to improve the performance and robustness of emotion classification systems based on audio data.</ns2:p>
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".