Brain-inspired Intelligence Technology

Brain-inspired Intelligence Technology

  • Developing a series of brain-inspired intelligence methods, with characteristics of good generalization, adaptability, robustness and interpretability.
  • Building an AI testing system in open environments, with database and testing protocols, supporting various AI tasks.
  • Founding an innovative research team with international influence.
  • Brain-inspired intelligence learning methods
  • Perception application with standards and strategies for comprehensive evaluation in open domain.
  • Differentiable Architecture Approximation (DATA) with an Ensemble Gumbel-Softmax (EGS) estimator.
  • Architecture parameters and network weights in the NAS model can be jointly optimized with the standard BP.

  • Exploiting convolutional neural network for point cloud processing is quite challenging, due to the inherent irregular distribution and discrete shape representation of point clouds.
  • A novel differential convolution search paradigm on point clouds, which works in a purely data-driven manner and thus is capable of auto-creating a group of suitable convolutions for geometric shape modeling.

  • It is the first time to highlight the generalized object recognition problem.
  • Propose a meta-leaning object recognition in unseen domains, combining gradient and meta-gradient in optimization.
  • Propose two benchmarks for open-set generalized object recognition evaluation.
  • Besides, we also use meta-learning to propose fast domain adaptation and discriminative image labeling way.

Learning Objective

  • Prototypes representing class distribution
  • Feature space: desired to be compact within each class
  • Classification Loss (CL):MCE, MCL (margin-based classification loss, GMCL, DCE (distance-based cross-entropy)

Compactness of each class benefits robustness to outlier (which has larger distance than within-class samples)