دانشگاه آزاد اسلامی واحد تهران جنوب

صفحه اصلی / نخستین همایش هوش مصنوعی و پژوهش های نوظهور: همگرایی انسان و سیستم های هوشمند

.Improving Knowledge Distillation Transfer Learning to Increase the Accuracy of the Lightweight Multi-Task Learning Classifier in Cognitive Software-Defined Radio

نویسندگان :

Ali Tohidi¹ Amir Amirabadi² Iman Ahanian³ Razieh Farazkish⁴

1- دانشگاه آزاد اسلامی واحد تهران جنوب 2- دانشگاه آزاد اسلامی واحد تهران جنوب 3- دانشگاه آزاد اسلامی واحد تهران جنوب 4- دانشگاه آزاد اسلامی واحد تهران جنوب

کلمات کلیدی :

Cognitive Software-Defined Radio،Lightweight neural networks

چکیده :

In the knowledge distillation transfer learning, due to the low capacity of the student model to capture transferred information gains from the teacher model, it is a challenge to increase the temperature coefficient (T) to achieve softer targets and more information gains. Also, inherent uncertainties in wireless communication data decrease the quality of the transmitted information. This paper proposes a solution to this challenge by applying active learning with fuzzy clustering sampling and fuzzy c-mean clustering on soft targets obtained from the temperature softmax activation function in an optimal temperature coefficient. This approach improves the quality of information transferred to the student model and, as a result, increases the accuracy of its performance. We implement the proposed learning approach on a lightweight classifier called the MobileNet V_3 MTL for simultaneous modulation and signal classification in cognitive software-defined radio (CSDR). The Fuzzy Active Knowledge Distillation (FAKD) learning method increases the model's generalizability. The proposed method is verified by applying new data received in the radio frequency spectrum, where a Raspberry Pi 4B with RTL-SDR is used to implement the design.