Denoised Internal Models:A Brain-inspired Autoencoder Against Adversarial Attacks

Despite its great success,deep learning severely suffers from robustness;i.e.,deep neural networks are very vulnerable to adversarial attacks,even the simplest ones.Inspired by recent advances in brain science,we propose the denoised internal models(DIM),a novel generative autoencoder-based model to tackle this challenge.Simulating the pipeline in the human brain for visual signal processing,DIM adopts a two-stage approach.In the first stage,DIM uses a denoiser to reduce the noise and the dimensions of inputs,reflecting the information pre-processing in the thalamus.Inspired by the sparse coding of memory-related traces in the primary visual cortex,the second stage produces a set of internal models,one for each category.We evaluate DIM over 42 adversarial attacks,showing that DIM effectively defenses against all the attacks and outperforms the SOTA on the overall robustness on the MNIST(Modified National Institute of Standards and Technology)dataset.

Epigenetic regulators sculpt the plastic brain

BACKGROUND： Epigenetic regulation is a level of transcriptional regulation that occurs in addition to the genetic programming found in biological systems. In the brain, the epigenetic machinery gives the system an opportunity to adapt to a given environment to help not only the individual but also the species survive and expand. However, such a regulatory system has risks, as mutations resulting from epigenetic regulation can cause severe neurological or psychiatric disorders. OBJECTIVE： Here, we review the most recent findings regarding the epigenetic mechanisms that control the activity- dependent gene transcription leading to synaptic plasticity and brain function and the defects in these mechanisms that lead to neurological disorders. METHODS： A search was carded out systematically, searching all relevant publications up to June 2017, using the PubMed search engine. The following keywords were used： ＂activity induced epigenetic,＂ ＂gene transcription,＂ and ＂neurological disorders.＂ RESULTS： A wide range of studies focused on the roles of epigenetics in transgenerational inheritance, neural differentiation, neural circuit assembly and brain diseases. Thirty-one articles focused specifically on activity-induced epigenetic modifications that regulated gene transcription and memory formation and consolidation. CONCLUSION： Activity-dependent epigenetic mechanisms of gene expression regulation contribute to basic neuronal physiology, and defects were associated with an elevated risk for brain disorders.

Do brain oscillations orchestrate memory?

Rhythmicity and oscillations are common features in nature, and can be seen in phenomena such as seasons, breathing, and brain activity. Despite the fact that a single neuron transmits its activity to its neighbor through a transient pulse, rhythmic activity emerges from large population-wide activity in the brain, and such rhythms are strongly coupled with the state and cognitive functions of the brain. However, it is still debated whether the oscillations of brain activity actually carry information. Here, we briefly introduce the biological findings of brain oscillations, and summarize the recent progress in understanding how oscillations mediate brain function. Finally, we examine the possible relationship between brain cognitive function and oscillation, focusing on how oscillation is related to memory, particularly with respect to state-dependent memory formation and memory retrieval under specific brain waves. We propose that oscillatory waves in the neocortex contribute to the synchronization and activation of specific memory trace ensembles in the neocortex by promoting long-range neural communication.