Uncertainty-Aware Multi-Task Learning for Joint Modulation Recognition and SINR Estimation

Kosar Nourolahi; Vahid Ghasemi

cs.NIAug 28, 2026

Uncertainty-Aware Multi-Task Learning for Joint Modulation Recognition and SINR Estimation

Authors: Kosar Nourolahi, Vahid Ghasemi

Organizations: Department of Artificial Intelligence, Kermanshah University of Technology, Kermanshah Province, Iran

Abstract

Joint modulation recognition and signal-to-interference-plus-noise ratio (SINR) estimation can reduce duplicated processing in intelligent receivers, but the two tasks have different uncertainty characteristics. This letter proposes an uncertainty-aware multi-task model that transforms each short normalized in-phase/quadrature window into 36 deterministic, label-free statistics, learns a shared representation, and uses task-specific adapters for modulation classification and heteroscedastic SINR regression. A joint uncertainty score combines classification entropy and predicted regression variance to support selective inference. Simulations cover QPSK, 8PSK, 16QAM, and 64QAM under matched additive white Gaussian noise/Rayleigh channels and an unseen frequency-selective Rician channel. Over five independent seeds, the proposed model improves matched and unseen-channel accuracy over conventional multi-task learning by 14.86 and 8.61 percentage points, respectively, while reducing SINR mean absolute error by 1.60 and 1.61 dB. Confidence-based rejection further lowers modulation error under channel mismatch.

Figures & tables

Panel of Figure 1 — Fig. 1: Proposed statistics-based uncertainty-aware multi-task inference chain.

Feature group	Dimension
I/Q/amplitude moments (orders 1-4)	12
Complex higher-order moments	5
Phase-increment harmonics (orders 1-4)	8
Lagged correlations ( $1,2,4,8$ )	8
Spectral entropy/flatness/concentration	3
Total	36

TABLE I: Deterministic input representation.

Test	Method	Accuracy	Macro-F1	MAE (dB)	RMSE (dB)
Matched	RBF-SVM/SVR	0.460	0.461	3.765	4.782
	Separate CNNs	$0.335\!\pm\!0.019$	$0.295\!\pm\!0.015$	$4.752\!\pm\!0.098$	$6.024\!\pm\!0.170$
	Conventional MTL	$0.352\!\pm\!0.009$	$0.297\!\pm\!0.022$	$4.904\!\pm\!0.303$	$6.310\!\pm\!0.439$
	Proposed	$\mathbf{0.500\!\pm\!0.004}$	$\mathbf{0.508\!\pm\!0.004}$	$\mathbf{3.302\!\pm\!0.062}$	$\mathbf{4.158\!\pm\!0.045}$
Unseen Rician	RBF-SVM/SVR	0.355	0.357	5.153	6.564
	Separate CNNs	$0.306\!\pm\!0.011$	$0.247\!\pm\!0.028$	$5.968\!\pm\!0.181$	$7.793\!\pm\!0.288$

TABLE II: Overall performance. Neural results are mean

\pm

standard deviation over five independent seeds.

Fig. 2: Matched-channel performance versus SINR: (a) modulation-classification accuracy and (b) SINR-estimation RMSE. Neural curves are means over five independent seeds.

Fig. 3: Selective inference under unseen Rician fading: (a) modulation-classification error and (b) SINR-estimation MAE versus retained coverage.

Fig. 4: Row-normalized confusion matrices of the proposed method under (a) matched channels and (b) the unseen frequency-selective Rician channel.

Model	Matched	Unseen Rician
Raw-I/Q hybrid	0.494/3.486	0.403/4.934
Final proposed	0.500/3.302	0.407/4.703

TABLE III: Validation-based architecture selection. Entries are accuracy/MAE (dB).

Automatic Modulation Classification

Automatic modulation classification identifies radio signal modulation schemes from received waveforms, often under noise, fading, and limited labeled data.

See all 18 papers

Joint Interference Detection and Identification via Adversarial Multi-task Learning
Apr 8, 2026H. Xu, L. Hu, B. He +1Automatic Modulation Classification Multi-Task Learning
College of Electronic Science and Technology, National University of Defense Technology, Changsha 410003, P. R. China
Deep Learning for Multi-Antenna Modulation Recognition of Radio Signals
Apr 20, 2026Tao Chen, Shilian Zheng, Jiepeng Chen +3Wireless Communications
College of Electrical and Information Engineering, Quzhou University, Quzhou 324000, China · National Key Laboratory of Electromagnetic Space Security, Jiaxing 314033, China · Institute of Cyberspace Security, and also with the College of Information Engineering, Zhejiang University of Technology, Hangzhou2026 310023, China
An Uncertainty-Driven Hybrid Deep Learning Approach for Broad-Coverage RF Modulation Recognition
Aug 1, 2026Nurettin Safak, Durdu Can Yerdeyatar, Muhammet Sefa Demirel +3Automatic Modulation Classification
Department of Electrical and Electronics Engineering, Ankara Yildirim Beyazit University, Ankara, Turkey

Apr 8, 2026J/K move · Enter open · S save

Joint Interference Detection and Identification via Adversarial Multi-task Learning

H. Xu, L. Hu, B. He, S. Wang

College of Electronic Science and Technology, National University of Defense Technology, Changsha 410003, P. R. China

Automatic Modulation Classification Multi-Task Learning Wireless Communications

Open paper AskPDF arXiv

Precise interference detection and identification are crucial for enhancing the survivability of communication systems in non-cooperative wireless environments. While deep learning (DL) has advanced this field, existing single-task learning (STL) approaches neglect inherent task correlations. Furthermore, emerging multi-task learning (MTL) methods often lack a theoretical foundation for quantifying and modeling task relationships. To bridge this gap, we establish a theoretically grounded MTL framework for joint interference detection, modulation identification, and interference identification. First, we derive an upper bound for the weighted expected loss in MTL frameworks. This bound explicitly connects MTL performance to task similarity, quantified by the Wasserstein distance and learnable task relation coefficients. Guided by this theory, we present the adversarial multi-task interference detection and identification network (AMTIDIN), which integrates adversarial training to minimize distributional discrepancies across tasks and uses adaptive coefficients to model task correlations dynamically. Crucially, we conducted a quantitative analysis of task similarity to reveal intrinsic task relationships, specifically that modulation identification and interference identification share a substantial feature overlap distinct from interference detection. Experiments demonstrate that AMTIDIN outperforms its independently trained single-task counterparts and MTL baselines under the evaluated conditions of limited training data, short signal lengths, and low signal-to-noise ratios (SNRs)

Abstract

Automatic Modulation Classification

Explore similar work

Joint Interference Detection and Identification via Adversarial Multi-task Learning

Deep Learning for Multi-Antenna Modulation Recognition of Radio Signals

An Uncertainty-Driven Hybrid Deep Learning Approach for Broad-Coverage RF Modulation Recognition