Aggregated Learning: A Deep Learning Framework Based on Information-Bottleneck Vector Quantization
This work addresses the problem of inefficient training in deep learning for researchers and practitioners, offering a novel framework that is incremental but with strong empirical gains.
The paper tackles the deficiency of standard neural networks as scalar quantizers by proposing AgrLearn, a framework based on vector information-bottleneck quantization that optimizes against multiple data samples, resulting in significant improvements in image recognition and text classification and reducing training samples by up to 80% for ResNet.
Based on the notion of information bottleneck (IB), we formulate a quantization problem called "IB quantization". We show that IB quantization is equivalent to learning based on the IB principle. Under this equivalence, the standard neural network models can be viewed as scalar (single sample) IB quantizers. It is known, from conventional rate-distortion theory, that scalar quantizers are inferior to vector (multi-sample) quantizers. Such a deficiency then inspires us to develop a novel learning framework, AgrLearn, that corresponds to vector IB quantizers for learning with neural networks. Unlike standard networks, AgrLearn simultaneously optimizes against multiple data samples. We experimentally verify that AgrLearn can result in significant improvements when applied to several current deep learning architectures for image recognition and text classification. We also empirically show that AgrLearn can reduce up to 80% of the training samples needed for ResNet training.