哪些人工智能技术可以申请专利? | 每日IP英文第435期
Training data preparation: collecting meaningful training data, balancing positive/negative samples in the training data, labeling the training data, standardization of the training data, encoding or embedding of the training data, synthetic training data generation. Novel machine learning architectures: new neural network architecture, hybrid model (e.g., a group of homogeneous neural networks working collectively, or a neural network trained based on training data from a general domain and subsequently transformed by training based on training data from specific domains), hierarchical model (e.g., federated learning). Loss function: a new loss function that improves training efficiency. Sparsification/Pruning of neural networks: reducing the number of active neurons in neural networks, reducing the number of channels/layers in neural networks. Output post-processing: converting predictions to probabilities when definitiveness is harmful.
大岭IP知识产权律师团队
赞 (0)