DeepNightLearnersUnderstanding Contrastive Representation Learning through Alignment and Uniformity on the HypersphereData can be represented as a vector in a continuous space, but often it is concentrated in a narrow area. Can it be improved?
DeepNightLearnersImproving Self-supervised Learning with Automated Unsupervised Outlier ArbitrationAugmenting images training data to add variability doesn't always end well since the augmentation changes the semantic content of the image. Here's how to solve it.
DeepNightLearnersUnsupervised Learning of Visual Features by Contrasting Cluster AssignmentsThis paper suggests a new computationally efficient method for constructing low-dimensional representation of unlabeled data.
Computer VisionNeRF: Representing Scenes as Neural Radiance Fields for View SynthesisA review and clear explanation of the NeRF method, which can be used to synthesize 3D scenes out of an input image. This method is the base for many other research methods that followed.
Computer VisionConvolutional Networks on Tabular dataYam Peleg examines a kaggle solution using convolutional neural networks which can process tabular data while being columns order agnostic.
Computer VisionPerceiver: General Perception with Iterative AttentionTo overcome Transformers' squared complexity (w.r.t input length), the Perceiver article here offers a novel method to learn the QKV matrices. Check it out!
DeepNightLearnersDiscriminator Rejection SamplingA summary of a method to improve GAN generated images by exploiting accumulated "information" from the Discriminator during the GAN training process.
DeepNightLearnersRepresentation learning via invariant causal mechanismsThis article suggests a method to construct augmentation resilient and styling invariant image representation in lower dimension (embeddings).
