MIM4DD: Mutual Information Maximization for Dataset Distillation Y uzhang Shang
–Neural Information Processing Systems
More details can be found in [10]. It allows us to transform the target problem at the data level (Eq. Given that each layer's mapping The ConvNet comprises three consecutive blocks of'Conv-InstNorm-ReLU-AvgPool.' The training is stopped after 5,000 iterations. To test the ConvNet's performance on the The network's initial learning rate is 0.01.
Neural Information Processing Systems
Oct-8-2025, 07:25:45 GMT