Deep Learning Through the Lens of Example Difficulty

By Robert J. N. Baldock, Hartmut Maennel, Behnam Neyshabur. 2021

200 citations recorded by the source corpus.

Existing work on understanding deep learning often employs measures that compress all data-dependent information into a few numbers. In this work, we adopt a perspective based on the role of individual examples. We introduce a measure of the computational difficulty of making a prediction for a given input: the (effective) prediction depth. Our extensive investigation reveals surprising yet simple relationships between the prediction depth of a given input and the model's uncertainty, confidence, accuracy and speed of learning for that data point. We further categorize difficult examples into three interpretable groups, demonstrate how these groups are processed differently inside deep models and showcase how this understanding allows us to improve prediction accuracy. Insights from our study lead to a coherent view of a number of separately reported phenomena in the literature: early layers generalize while later layers memorize; early layers converge faster and networks learn easy data and simple functions first.

Subfields: Artificial Intelligence

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

Task-Aware Information Routing from Common Representation Space in Lifelong Learning (2023, background): “The intuition behind placing TAMs higher up in the layer hierarchy is as follows: the early layers of DNNs capture generic information, while the later layers memorize due to the diminishing dimension and radius of the manifold (Stephenson et al., 2020; Baldock et al., 2021).”
Deep Learning on a Data Diet: Finding Important Examples Early in Training (2021, background): “Another quite different approach estimates example difficulty using prediction depth [34], which is defined as the first layer at which a k-Nearest Neighbor classifier can correctly classify an example using the representation of the image in all subsequent layers.”
Interpretable deep learning: interpretation, interpretability, trustworthiness, and beyond (2021, methodology): “– To highlight the important parts of input features on which the deep model mainly relies, using gradients [155], perturbations [55], proxy explainable models [137] and other methods; – To investigate the inside of deep models to understand the rationale of how models make decisions by visualizing ”
Eliciting Latent Predictions from Transformers with the Tuned Lens (2023, unknown): “Following Baldock et al. (2021)’s work on computer vision models, we define the prediction depth of a prompt x to be the number of layers after which a model’s top-1 prediction for x stops changing.”
Metadata Archaeology: Unearthing Data Subsets by Leveraging Training Dynamics (2022, methodology): “Prior work has developed a rough taxonomy of data properties, or metadata which different examples might exhibit, including but not limited to: noisy (Wu et al., 2020; Yi and Wu, 2019; Thulasidasan et al., 2019a;b), atypical (Hooker et al., 2020; Buolamwini and Gebru, 2018; Hashimoto et al., 2018;…”