… the advantages of contrastive learning, we propose probabilistic … Probabilistic Representations
for Video Contrastive Learning (ProViCo), in which videos are represented as probability …

… If we sample temporally distant clips with smaller probabilities, the contrastive loss (Equation
1) would focus more on the temporally close clips, pulling their features closer and …

… has been shown to produce generalizable representations of audio and visual data by …
representations for downstream tasks. In this paper, we propose an active contrastive learning …

… use contrastive learning to preserve video-text semantic relatedness. By maximizing the joint
posterior probability of video and text, we find a semantically related subspace for compact …

… for selfsupervised representation learning on videos. This … future representations; Second,
we propose a curriculum training … , contrast, saturation, hue and random greyscale during …

… We evaluate the effect of our probabilistic targeted sampling on the final performance. We
observe a performance boost with targeted sampling (temperature β = 5) compared to uniform …

… Next, the context c is passed to a spatial pooling layer to get a feature vector, and then a
fully-connected layer and a multi-class softmax function outputs the probabilities for video action …

… videos with multiple viewpoints as in Time-Contrastive … property of video and applies contrastive
learning on a sequence … to learn a language model using a probabilistic contrastive loss, …

… video representation learning based on Contrastive Predictive Coding (CPC) [27]. Previously,
CPC has been used to learn representations … and contrastive estimation to learn long-term …

… In contrast, we focus on learning the motion patterns within … Therefore the CNN returns the
conditional probability of action … log probability of the class y given video X is obtained by ∑m …