Reducing Overlearning through Disentangled Representations by Suppressing Unknown Tasks

TL;DR

This paper introduces a model-agnostic method to reduce overlearning in deep visual models by suppressing unknown tasks, along with a new trust score metric and a benchmark dataset for evaluation.

Contribution

It presents a novel approach to suppress all unknown tasks in deep models without requiring ground truth labels, and introduces a trust score metric and a benchmark dataset.

Findings

Effective suppression of overlearning in multiple models

Improved trust scores indicating better model reliability

Successful application on various datasets including face attributes

Abstract

Existing deep learning approaches for learning visual features tend to overlearn and extract more information than what is required for the task at hand. From a privacy preservation perspective, the input visual information is not protected from the model; enabling the model to become more intelligent than it is trained to be. Current approaches for suppressing additional task learning assume the presence of ground truth labels for the tasks to be suppressed during training time. In this research, we propose a three-fold novel contribution: (i) a model-agnostic solution for reducing model overlearning by suppressing all the unknown tasks, (ii) a novel metric to measure the trust score of a trained deep learning model, and (iii) a simulated benchmark dataset, PreserveTask, having five different fundamental image classification tasks to study the generalization nature of models. In the first set of experiments, we learn disentangled representations and suppress overlearning of five popular deep learning models: VGG16, VGG19, Inception-v1, MobileNet, and DenseNet on PreserverTask dataset. Additionally, we show results of our framework on color-MNIST dataset and practical applications of face attribute preservation in Diversity in Faces (DiF) and IMDB-Wiki dataset.