One of the most overlooked challenges in dance generation is that the autoregressiveframeworks are prone tofreezing motions due tonoiseaccumulation. Inthispaper,wepresent twomodules thatcanbeplugged intotheexisting models to enable them to generate non-freezing and high fidelity dances.

The majority of existing multimodal sequential learning methods focus on how to obtain powerful individual representations and neglect to effectively capture themultimodal joint representation. Bilinear attention network (BAN) isacommonly used integration method, which leverages tensor operations to associate thefeatures ofdifferent modalities.

Personalized federated learning (PFL) has been widely investigated to address the challenge of data heterogeneity, especially when a single generic model is inadequate in satisfying the diverse performance requirements of local clients simultaneously.

Given training data ofn-dimensional LPs, we learn ann k projection matrix withn>k.