Learning Decomposable and Debiased Representations via Attribute-Centric Information Bottlenecks

Biased attributes, spuriously correlated with target labels in a dataset, can problematically lead to neural networks that learn improper shortcuts for classifications and limit their capabilities for out-of-distribution (OOD) generalization. Although many debiasing approaches have been proposed to ensure correct predictions from biased datasets, few studies have considered learning latent embedding consisting of intrinsic and biased attributes that contribute to improved performance and explain how the model pays attention to attributes. In this paper, we propose a novel debiasing framework, Debiasing Global Workspace, introducing attention-based information bottlenecks for learning compositional representations of attributes without defining specific bias types. Based on our observation that learning shape-centric representation helps robust performance on OOD datasets, we adopt those abilities to learn robust and generalizable representations of decomposable latent embeddings corresponding to intrinsic and biasing attributes. We conduct comprehensive evaluations on biased datasets, along with both quantitative and qualitative analyses, to showcase our approach's efficacy in attribute-centric representation learning and its ability to differentiate between intrinsic and bias-related features.

RNAS-CL: Robust Neural Architecture Search by Cross-Layer Knowledge Distillation

Deep Neural Networks are vulnerable to adversarial attacks. Neural Architecture Search (NAS), one of the driving tools of deep neural networks, demonstrates superior performance in prediction accuracy in various machine learning applications. However, it is unclear how it performs against adversarial attacks. Given the presence of a robust teacher, it would be interesting to investigate if NAS would produce robust neural architecture by inheriting robustness from the teacher. In this paper, we propose Robust Neural Architecture Search by Cross-Layer Knowledge Distillation (RNAS-CL), a novel NAS algorithm that improves the robustness of NAS by learning from a robust teacher through cross-layer knowledge distillation. Unlike previous knowledge distillation methods that encourage close student/teacher output only in the last layer, RNAS-CL automatically searches for the best teacher layer to supervise each student layer. Experimental result evidences the effectiveness of RNAS-CL and shows that RNAS-CL produces small and robust neural architecture.

Similarity-based Distance for Categorical Clustering using Space Structure

Clustering is spotting pattern in a group of objects and resultantly grouping the similar objects together. Objects have attributes which are not always numerical, sometimes attributes have domain or categories to which they could belong to. Such data is called categorical data. To group categorical data many clustering algorithms are used, among which k- modes algorithm has so far given the most significant results. Nevertheless, there is still a lot which could be improved. Algorithms like k-means, fuzzy-c-means or hierarchical have given far better accuracies with numerical data. In this paper, we have proposed a novel distance metric, similarity-based distance (SBD) to find the distance between objects of categorical data. Experiments have shown that our proposed distance (SBD), when used with the SBC (space structure based clustering) type algorithm significantly outperforms the existing algorithms like k-modes or other SBC type algorithms when used on categorical datasets.

Better Together: Resnet-50 accuracy with $13x$ fewer parameters and at $3x$ speed

Recent research on compressing deep neural networks has focused on reducing the number of parameters. Smaller networks are easier to export and deploy on edge-devices. We introduce Adjoined networks as a training approach that can compress and regularize any CNN-based neural architecture. Our one-shot learning paradigm trains both the original and the smaller networks together. The parameters of the smaller network are shared across both the architectures. For resnet-50 trained on Imagenet, we are able to achieve a $13.7x$ reduction in the number of parameters and a $3x$ improvement in inference time without any significant drop in accuracy. For the same architecture on CIFAR-100, we are able to achieve a $99.7x$ reduction in the number of parameters and a $5x$ improvement in inference time. On both these datasets, the original network trained in the adjoint fashion gains about $3\%$ in top-1 accuracy as compared to the same network trained in the standard fashion.