In neural network compression, most current methods reduce unnecessary parameters by measuring importance and redundancy. To augment already highly optimized existing solutions, we propose linearity-based compression as a novel way to reduce weights in a neural network. It is based on the intuition that with ReLU-like activation functions, neurons that are almost always activated behave linearly, allowing for merging of subsequent layers. We introduce the theory underlying this compression and evaluate our approach experimentally. Our novel method achieves a lossless compression down to 1/4 of the original model size in over the majority of tested models. Applying our method on already importance-based pruned models shows very little interference between different types of compression, demonstrating the option of successful combination of techniques. Overall, our work lays the foundation for a new type of compression method that enables smaller and ultimately more efficient neural network models.

The Yongle Palace murals, as valuable cultural heritage, have suffered varying degrees of damage, making their restoration of significant importance. However, the giant size and unique data of Yongle Palace murals present challenges for existing deep-learning based restoration methods: 1) The distinctive style introduces domain bias in traditional transfer learning-based restoration methods, while the scarcity of mural data further limits the applicability of these methods. 2) Additionally, the giant size of these murals results in a wider range of defect types and sizes, necessitating models with greater adaptability. Consequently, there is a lack of focus on deep learning-based restoration methods for the unique giant murals of Yongle Palace. Here, a 3M-Hybrid model is proposed to address these challenges. Firstly, based on the characteristic that the mural data frequency is prominent in the distribution of low and high frequency features, high and low frequency features are separately abstracted for complementary learning. Furthermore, we integrate a pre-trained Vision Transformer model (VIT) into the CNN module, allowing us to leverage the benefits of a large model while mitigating domain bias. Secondly, we mitigate seam and structural distortion issues resulting from the restoration of large defects by employing a multi-scale and multi-perspective strategy, including data segmentation and fusion. Experimental results demonstrate the efficacy of our proposed model. In regular-sized mural restoration, it improves SSIM and PSNR by 14.61% and 4.73%, respectively, compared to the best model among four representative CNN models. Additionally, it achieves favorable results in the final restoration of giant murals.

AMSTERDAM, June 23 (Reuters) - For the first time in 300 years, Rembrandt's famed "The Night Watch" is back on display in what researchers say is its original size, with missing parts temporarily restored in an exhibition aided by artificial intelligence. Rembrandt finished the large canvas, which portrays the captain of an Amsterdam city militia ordering his men into action, in 1642. Although it is now considered one of the greatest masterpieces of the Dutch Golden Age, strips were cut from all four sides of it during a move in 1715. Though those strips have not been found, another artist of the time had made a copy, and restorers and computer scientists have used that, blended with Rembrandt's style, to recreate the missing parts. "It's never the real thing, but I think it gives you different insight into the composition," Rijksmuseum director Taco Dibbits said.

MIT researchers have created a real life'shrink ray' that can reduce 3D structures to one thousandth of their original size. Scientists can put all kinds of useful materials in the polymer before they shrink it, including metals, quantum dots, and DNA. The process is essentially the opposite of expansion microscopy, which is widely used by scientists to create 3D visualisations of microscopic cells. Instead of making things bigger, scientists attach special molecules which block negative charges between molecules so they no longer repel which makes them contract. Experts say that making such tiny structures could be useful in many fields, including in medicine and for creating nanoscale robotics.

An autoencoder has two parts: an encoder and a decoder. The encoder reduces the dimensions of input data so that the original information is compressed. The decoder restores the original information from the compressed data. The autoencoder is a neural network that learns to encode and decode automatically (hence, the name). Once learning is done, we can use the encoder and decoder independently.

Artificial muscles inspired by the Japanese folding technique of origami could give robots the power to lift up to 1,000 times their own weight. US researchers have crafted a cheap new material that will let the machines carry out smoother, less rigid, and more human-like movements. The advance offers a leap forward in the field of soft robotics, which are fast replacing older generations of automatons. Researchers at the Massachusetts Institute of Technology have crafted cheap, artificial muscles for robots that give them the power to lift up to 1,000 times their own weight. Researchers built dozens of muscles, using metal springs, packing foam or plastic in a range of shapes and sizes.

Similar to convolution neural networks, recurrent neural networks (RNNs) typically suffer from over-parameterization. Quantizing bit-widths of weights and activations results in runtime efficiency on hardware, yet it often comes at the cost of reduced accuracy. This paper proposes a quantization approach that increases model size with bit-width reduction. This approach will allow networks to perform at their baseline accuracy while still maintaining the benefits of reduced precision and overall model size reduction.