Semi-supervised tuning from temporal coherence

By Davide Maltoni, Vincenzo Lomonaco. 2015

17 citations recorded by the source corpus.

Recent works demonstrated the usefulness of temporal coherence to regularize supervised training or to learn invariant features with deep architectures. In particular, enforcing a smooth output change while presenting temporally-closed frames from video sequences, proved to be an effective strategy. In this paper we prove the efficacy of temporal coherence for semi-supervised incremental tuning. We show that a deep architecture, just mildly trained in a supervised manner, can progressively improve its classification accuracy, if exposed to video sequences of unlabeled data. The extent to which, in some cases, a semi-supervised tuning allows to improve classification accuracy (approaching the supervised one) is somewhat surprising. A number of control experiments pointed out the fundamental role of temporal coherence.

How researchers see this paper

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

Citations in the corpus

Continuous Learning in Single-Incremental-Task Scenarios (2018, background): “e realistic scenario for real applications, and would constitute the main target of our future research. AR1 extension to unsupervised (or semi-supervised) implementations, such as those described in [22] and [26] is another scenario of interest for future studies. In particular, the 2-level self-or”
CORe50: a New Dataset and Benchmark for Continuous Object Recognition (2017, background): “Similar ﬁndings have been reported for NORB , COIL100 and BigBrother in the NI scenario [16] [15].”
Deep Predictive Coding Networks for Video Prediction and Unsupervised Learning (2016, methodology): “While not rivaling supervised methods, this approach, and others that rely on temporal coherence, have achieved some measure of success, pointing to the potential of learning useful representations from video [15, 27, 30, 50, 55].”
Rehearsal-Free Continual Learning over Small Non-I.I.D. Batches (2019, methodology): “Inspired by previous finding with Hierarchical Temporal Memories [31, 21] where gradient descent by HSR only affects coincidence pooling, here we propose to fine-tune DWSC architectures by freezing depthwise spatial filters and leaving pointwise poolers free to learn.”
Fine-Grained Continual Learning (2019, methodology): “Inspired by previous ﬁnding with Hierarchical Temporal Memories [23, 24] where gradient descent by HSR only affects coincidence pooling, here we propose to ﬁne-tune DWSC architectures by freezing depthwise spatial ﬁlters and leaving pointwise poolers free to learn.”
Online Continual Learning on Sequences (2020, background): “For NORB and COIL-100 we defined in [46] a number of exploration sequences that turn the native static benchmarks into continual learning tasks; [46] also reports supervised and semi-supervised accuracy for the NI scenario.”