• Corpus ID: 251320207

On the Learnability of Physical Concepts: Can a Neural Network Understand What's Real?

  title={On the Learnability of Physical Concepts: Can a Neural Network Understand What's Real?},
  author={Alessandro Achille and Stefan 0 Soatto},
We revisit the classic signal-to-symbol barrier in light of the remarkable ability of deep neural networks to generate realistic synthetic data. DeepFakes and spoofing highlight the feebleness of the link between physical reality and its abstract representation, whether learned by a digital computer or a biological agent. Starting from a widely applicable definition of abstract concept , we show that standard feed-forward architectures cannot capture but trivial concepts, regardless of the number… 

Figures from this paper



The Information Complexity of Learning Tasks, their Structure and their Distance

This work introduces an asymmetric distance in the space of learning tasks, and a framework to compute their complexity, the first to measure complexity in a way that accounts for the effect of the optimization scheme, which is critical in Deep Learning.

Attention is All you Need

A new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely is proposed, which generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data.

Attention is Turing-Complete

It is shown that the Transformer with hard-attention is Turing complete exclusively based on their capacity to compute and access internal dense representations of the data.

Generating Text with Recurrent Neural Networks

The power of RNNs trained with the new Hessian-Free optimizer by applying them to character-level language modeling tasks is demonstrated, and a new RNN variant that uses multiplicative connections which allow the current input character to determine the transition matrix from one hidden state vector to the next is introduced.

On the difficulty of training recurrent neural networks

This paper proposes a gradient norm clipping strategy to deal with exploding gradients and a soft constraint for the vanishing gradients problem and validates empirically the hypothesis and proposed solutions.

Transformers are RNNs: Fast Autoregressive Transformers with Linear Attention

This work expresses the self-attention as a linear dot-product of kernel feature maps and makes use of the associativity property of matrix products to reduce the complexity from O(N) to N, where N is the sequence length.

NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis

This work describes how to effectively optimize neural radiance fields to render photorealistic novel views of scenes with complicated geometry and appearance, and demonstrates results that outperform prior work on neural rendering and view synthesis.

Signal-to-symbol transformation and vice versa: from fundamental processes to representation

It is believed that to understand knowledge representation fully the authors must understand this transformation from signal to symbol and vice versa.

Convolutional neural networks applied to house numbers digit classification

This work augmented the traditional ConvNet architecture by learning multi-stage features and by using Lp pooling and establishes a new state-of-the-art of 95.10% accuracy on the SVHN dataset (48% error improvement).

On the set of images modulo viewpoint and contrast changes

It is shown that one can compute deterministic functions of the image that contain all the "information" present in the original image, except for the effects of viewpoint and illumination, and thus the " information" in an image that is relevant for recognition is sparse.