Mining Data Impressions from Deep Models as Substitute for the Unavailable Training Data


Deep models that have been through training retain their acquired knowledge in the form of model parameters. These parameters serve as "memory" for the trained models, which enables them to generalize well on data that they have not previously encountered. The utility of a trained model, on the other hand, is severely limited in the absence of training data, and can only be used for either inference or better initialization towards a target task. In this paper, we take it a step further and extract synthetic data by making use of the parameters that were learned about the model. These representations of the training data, which we refer to as Data Impressions, can be implemented in a variety of contexts and serve as proxies for the original data. These are helpful in situations in which only the pretrained models are available, and the training data is not being distributed (e.g., due to privacy or sensitivity concerns). We demonstrate how data impressions can be used to solve a variety of problems in computer vision, including unsupervised domain adaptation, continuous learning, and knowledge distillation. In addition to this, we investigate the robustness of lightweight models to adversarial inputs after being trained through knowledge distillation with these data impressions. In addition, we show that data impressions are effective in the generation of data-free Universal Adversarial Perturbations (UAPs) that have higher fooling rates. Extensive experiments carried out on benchmark datasets demonstrate that competitive performance can be achieved using data impressions even in the absence of the original training data.

Did you like this research project?

To get this research project Guidelines, Training and Code... Click Here

PROJECT TITLE : Stat-DSM: Statistically Discriminative Sub-Trajectory Mining With Multiple Testing Correction ABSTRACT: We propose a novel statistical approach, which we call Statistically Discriminative Sub-trajectory Mining
PROJECT TITLE : Periodic Communities Mining in Temporal Networks Concepts and Algorithms ABSTRACT: The occurrence of the phenomenon known as periodicity in social interactions within temporal networks is fairly common. Understanding
PROJECT TITLE : Mining High Quality Patterns Using Multi-Objective Evolutionary Algorithm ABSTRACT: The term "pattern mining," or PM for short, refers to the process of extracting from data patterns that are of interest to users.
PROJECT TITLE : Semantics of Data Mining Services in Cloud Computing ABSTRACT: Users now have access to extremely comprehensive data analysis tools that include all of the benefits that come along with working in an environment
PROJECT TITLE : Traffic Information Mining From Social Media Based on the MC-LSTM-Conv Model ABSTRACT: The use of social media such as Sina Weibo has the advantage of reflecting traffic information. This information can include

Ready to Complete Your Academic MTech Project Work In Affordable Price ?

Project Enquiry