Transfer learning is the predominant method for adapting pre-trained models on another task to new domains while preserving their internal architectures and augmenting them with requisite layers in Deep Neural Network models. Training intricate pre-trained models on a sizable dataset requires significant resources to fine-tune hyperparameters carefully. Most existing initialization methods mainly focus on gradient flow-related problems, such as gradient vanishing or exploding, or other existing approaches that require extra models that do not consider our setting, which is more practical. To address these problems, we suggest employing gradient-free heuristic methods to initialize the weights of the final new-added fully connected layer in neural networks from a small set of training data with fewer classes. The approach relies on partitioning the output values from pre-trained models for a small set into two separate intervals determined by the targets. This process is framed as an optimization problem for each output neuron and class. The optimization selects the highest values as weights, considering their direction towards the respective classes. Furthermore, empirical 145 experiments involve a variety of neural network models tested across multiple benchmarks and domains, occasionally yielding accuracies comparable to those achieved with gradient descent methods by using only small subsets.

Evaluating artificial intelligence (AI) systems is crucial for their successful deployment and safe operation in real-world applications. The assessor meta-learning model has been recently introduced to assess AI system behaviors developed from emergent characteristics of AI systems and their responses on a test set. The original approach lacks covering continuous ranges, for example, regression problems, and it produces only the probability of success. In this work, to address existing limitations and enhance practical applicability, we propose an assessor feedback mechanism designed to identify and learn from AI system errors, enabling the system to perform the target task more effectively while concurrently correcting its mistakes. Our empirical analysis demonstrates the efficacy of this approach. Specifically, we introduce a transition methodology that converts prediction errors into relative success, which is particularly beneficial for regression tasks. We then apply this framework to both neural network and support vector machine models across regression and classification tasks, thoroughly testing its performance on a comprehensive suite of 30 diverse datasets. Our findings highlight the robustness and adaptability of the assessor feedback mechanism, showcasing its potential to improve model accuracy and reliability across varied data contexts.