In this work, we propose Mutual Information Maximization Knowledge Distillation (MIMKD). Our method uses a contrastive objective to simultaneously estimate and maximize a lower bound on the mutual information of local and global feature representations between a teacher and a student network. We demonstrate through extensive experiments that this can be used to improve the performance of low capacity models by transferring knowledge from more performant but computationally expensive models. This can be used to produce better models that can be run on devices with low computational resources. Our method is flexible, we can distill knowledge from teachers with arbitrary network architectures to arbitrary student networks. Our empirical results show that MIMKD outperforms competing approaches across a wide range of student-teacher pairs with different capacities, with different architectures, and when student networks are with extremely low capacity. We are able to obtain 74.55% accuracy on CIFAR100 with a ShufflenetV2 from a baseline accuracy of 69.8% by distilling knowledge from ResNet-50. On Imagenet we improve a ResNet-18 network from 68.88% to 70.32% accuracy (1.44%+) using a ResNet-34 teacher network.
more »
« less
Improving Efficient Semantic Segmentation Networks by Enhancing Multi-scale Feature Representation via Resolution Path Based Knowledge Distillation and Pixel Shuffle
Multi-resolution paths and multi-scale feature representation are key elements of semantic segmentation networks. We develop two techniques for efficient networks based on the recent FasterSeg network architecture. One is to use a state-of-the-art high resolution network (e.g. HRNet) as a teacher to distill a light weight student network. Due to dissimilar structures in the teacher and student networks, distillation is not effective to be carried out directly in a standard way. To solve this problem, we introduce a tutor network with an added high resolution path to help distill a student network which improves FasterSeg student while maintaining its parameter/FLOPs counts. The other finding is to replace standard bilinear interpolation in the upscaling module of FasterSeg student net by a depth-wise separable convolution and a Pixel Shuffle module which leads to 1.9% (1.4%) mIoU improvements on low (high) input image sizes without increasing model size. A combination of these techniques will be pursued in future works.
more »
« less
- NSF-PAR ID:
- 10335366
- Date Published:
- Journal Name:
- Lecture notes in computer science
- Volume:
- 13018
- ISSN:
- 0302-9743
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
In this paper, we present a simple approach to train Generative Adversarial Networks (GANs) in order to avoid a mode collapse issue. Implicit models such as GANs tend to generate better samples compared to explicit models that are trained on tractable data likelihood. However, GANs overlook the explicit data density characteristics which leads to undesirable quantitative evaluations and mode collapse. To bridge this gap, we propose a hybrid generative adversarial network (HGAN) for which we can enforce data density estimation via an autoregressive model and support both adversarial and likelihood framework in a joint training manner which diversify the estimated density in order to cover different modes. We propose to use an adversarial network to transfer knowledge from an autoregressive model (teacher) to the generator (student) of a GAN model. A novel deep architecture within the GAN formulation is developed to adversarially distill the autoregressive model information in addition to simple GAN training approach. We conduct extensive experiments on real-world datasets (i.e., MNIST, CIFAR-10, STL-10) to demonstrate the effectiveness of the proposed HGAN under qualitative and quantitative evaluations. The experimental results show the superiority and competitiveness of our method compared to the baselines.more » « less
-
Recent deep-learning models have achieved impressive prediction performance, but often sacrifice interpretability and computational efficiency. Interpretability is crucial in many disciplines, such as science and medicine, where models must be carefully vetted or where interpretation is the goal itself. Moreover, interpretable models are concise and often yield computational efficiency. Here, we propose adaptive wavelet distillation (AWD), a method which aims to distill information from a trained neural network into a wavelet transform. Specifically, AWD penalizes feature attributions of a neural network in the wavelet domain to learn an effective multi-resolution wavelet transform. The resulting model is highly predictive, concise, computationally efficient, and has properties (such as a multi-scale structure) which make it easy to interpret. In close collaboration with domain experts, we showcase how AWD addresses challenges in two real-world settings: cosmological parameter inference and molecular-partner prediction. In both cases, AWD yields a scientifically interpretable and concise model which gives predictive performance better than state-of-the-art neural networks. Moreover, AWD identifies predictive features that are scientifically meaningful in the context of respective domains. All code and models are released in a full-fledged package available on Github.more » « less
-
Recently, contrastive learning has achieved great results in self-supervised learning, where the main idea is to pull two augmentations of an image (positive pairs) closer compared to other random images (negative pairs). We argue that not all negative images are equally negative. Hence, we introduce a self-supervised learning algorithm where we use a soft similarity for the negative images rather than a binary distinction between positive and negative pairs. We iteratively distill a slowly evolving teacher model to the student model by capturing the similarity of a query image to some random images and transferring that knowledge to the student. Specifically, our method should handle unbalanced and unlabeled data better than existing contrastive learning methods, because the randomly chosen negative set might include many samples that are semantically similar to the query image. In this case, our method labels them as highly similar while standard contrastive methods label them as negatives. Our method achieves comparable results to the state-of-the-art models.more » « less
-
more » « less
We present a tomographic imaging technique, termed Deep Prior Diffraction Tomography (DP-DT), to reconstruct the 3D refractive index (RI) of thick biological samples at high resolution from a sequence of low-resolution images collected under angularly varying illumination. DP-DT processes the multi-angle data using a phase retrieval algorithm that is extended by a deep image prior (DIP), which reparameterizes the 3D sample reconstruction with an untrained, deep generative 3D convolutional neural network (CNN). We show that DP-DT effectively addresses the missing cone problem, which otherwise degrades the resolution and quality of standard 3D reconstruction algorithms. As DP-DT does not require pre-captured data or pre-training, it is not biased towards any particular dataset. Hence, it is a general technique that can be applied to a wide variety of 3D samples, including scenarios in which large datasets for supervised training would be infeasible or expensive. We applied DP-DT to obtain 3D RI maps of bead phantoms and complex biological specimens, both in simulation and experiment, and show that DP-DT produces higher-quality results than standard regularization techniques. We further demonstrate the generality of DP-DT, using two different scattering models, the first Born and multi-slice models. Our results point to the potential benefits of DP-DT for other 3D imaging modalities, including X-ray computed tomography, magnetic resonance imaging, and electron microscopy.
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- https://doi.org/10.1007/978-3-030-90436-4_26
