We propose a novel Multi-Task Learning with Low Rank Attribute Embedding (MTL-LORAE) framework for person re-identification. Re-identifications from multiple cameras are regarded as related tasks to exploit shared information to improve re-identification accuracy. Both low level features and semantic/data-driven attributes are utilized. Since attributes are generally correlated, we introduce a low rank attribute embedding into the MTL formulation to embed original binary attributes to a continuous attribute space, where incorrect and incomplete attributes are rectified and recovered to better describe people. The learning objective function consists of a quadratic loss regarding class labels and an attribute embedding error, which is solved by an alternating optimization procedure. Experiments on four person re-identification datasets have demonstrated that MTL-LORAE outperforms existing approaches by a large margin and produces promising results.

Many of the genes disrupted in autism are identified as histone-modifying enzymes and chromatin remodelers, most prominently those that mediate histone methylation/demethylation. However, the role of histone methylation enzymes in the pathophysiology and treatment of autism remains unknown. To address this, we used mouse models of haploinsufficiency of the Shank3 gene (a highly penetrant monogenic autism risk factor), which exhibits prominent autism-like social deficits. We found that histone methyltransferases EHMT1 and EHMT2, as well as histone lysine 9 dimethylation (specifically catalyzed by EHMT1/2), were selectively increased in the prefrontal cortex (PFC) of Shank3-deficient mice and autistic human postmortem brains. Treatment with the EHMT1/2 inhibitor UNC0642 or knockdown of EHMT1/2 in PFC induced a robust rescue of autism-like social deficits in Shank3-deficient mice, and restored NMDAR-mediated synaptic function. Activity-regulated cytoskeleton-associated protein (Arc) was identified as one of the causal factors underlying the rescuing effects of UNC0642 on NMDAR function and social behaviors in Shank3-deficient mice. UNC0642 treatment also restored a large set of genes involved in neural signaling in PFC of Shank3-deficient mice. These results suggest that targeting histone methylation enzymes to adjust gene expression and ameliorate synaptic defects could be a potential therapeutic strategy for autism.

We propose Multi-Task Learning with Low Rank Attribute Embedding (MTL-LORAE) to address the problem of person re-identification on multi-cameras. Re-identifications on different cameras are considered as related tasks, which allows the shared information among different tasks to be explored to improve the re-identification accuracy. The MTL-LORAE framework integrates low-level features with mid-level attributes as the descriptions for persons. To improve the accuracy of such description, we introduce the low-rank attribute embedding, which maps original binary attributes into a continuous space utilizing the correlative relationship between each pair of attributes. In this way, inaccurate attributes are rectified and missing attributes are recovered. The resulting objective function is constructed with an attribute embedding error and a quadratic loss concerning class labels. It is solved by an alternating optimization strategy. The proposed MTL-LORAE is tested on four datasets and is validated to outperform the existing methods with significant margins.

The human 16p11.2 gene locus is a hot-spot for copy number variations which predispose carriers to a range of neuropsychiatric phenotypes. Microduplications of 16p11.2 are associated with autism spectrum disorder (ASD), intellectual disability (ID) and schizophrenia (SZ). Despite the debilitating nature of 16p11.2 duplications, the underlying molecular mechanisms remain poorly understood. Here we performed a comprehensive behavioral characterization of 16p11.2 duplication mice (16p11.2 dp/+ ) and identified social and cognitive deficits reminiscent of ASD and ID phenotypes. 16p11.2 dp/+ mice did not exhibit the SZ-related sensorimotor gating deficits, psychostimulant-induced hypersensitivity or motor impairment. Electrophysiological recordings of 16p11.2 dp/+ mice found the deficient GABAergic synaptic transmission and elevated neuronal excitability in the prefrontal cortex (PFC), a brain region critical for social and cognitive functions. RNA-sequencing identified genome-wide transcriptional aberrance in the PFC of 16p11.2 dp/+ mice, including downregulation of the GABA synapse regulator Npas4. Restoring Npas4 expression in PFC of 16p11.2 dp/+ mice ameliorated the social and cognitive deficits and reversed the GABAergic synaptic impairment and neuronal hyper-excitability. These findings suggest that prefrontal cortical GABAergic synaptic circuitry and Npas4 are strongly implicated in 16p11.2 duplication pathology, and may represent potential targets for therapeutic intervention in ASD.

Epigenetic aberration is implicated in aging and neurodegeneration. Using postmortem tissues from patients with Alzheimer’s disease (AD) and AD mouse models, we have found that the permissive histone mark H3K4me3 and its catalyzing enzymes are significantly elevated in the prefrontal cortex (PFC). Inhibiting H3K4-specific methyltransferases with the compound WDR5-0103 leads to the substantial recovery of PFC synaptic function and memory-related behaviors in AD mice. Among the up-regulated genes reversed by WDR5-0103 treatment in PFC of AD mice, many have the increased H3K4me3 enrichment at their promoters. One of the identified top-ranking target genes, Sgk1, which encodes serum and glucocorticoid-regulated kinase 1, is also significantly elevated in PFC of patients with AD. Administration of a specific Sgk1 inhibitor reduces hyperphosphorylated tau protein, restores PFC glutamatergic synaptic function, and ameliorates memory deficits in AD mice. These results have found a novel epigenetic mechanism and a potential therapeutic strategy for AD and related neurodegenerative disorders.

Machine learning is essentially the sciences of playing with data. An adaptive data selection strategy, enabling to dynamically choose different data at various training stages, can reach a more effective model in a more efficient way. In this paper, we propose a deep reinforcement learning framework, which we call Neural Data Filter (NDF), to explore automatic and adaptive data selection in the training process. In particular, NDF takes advantage of a deep neural network to adaptively select and filter important data instances from a sequential stream of training data, such that the future accumulative reward (e.g., the convergence speed) is maximized. In contrast to previous studies in data selection that is mainly based on heuristic strategies, NDF is quite generic and thus can be widely suitable for many machine learning tasks. Taking neural network training with stochastic gradient descent (SGD) as an example, comprehensive experiments with respect to various neural network modeling (e.g., multi-layer perceptron networks, convolutional neural networks and recurrent neural networks) and several applications (e.g., image classification and text understanding) demonstrate that NDF powered SGD can achieve comparable accuracy with standard SGD process by using less data and fewer iterations.

We Microsoft Research Asia made submissions to 11 language directions in the WMT19 news translation tasks. We won the first place for 8 of the 11 directions and the second place for the other three. Our basic systems are built on Transformer, back translation and knowledge distillation. We integrate several of our rececent techniques to enhance the baseline systems: multi-agent dual learning (MADL), masked sequence-to-sequence pre-training (MASS), neural architecture optimization (NAO), and soft contextual data augmentation (SCA).

No abstract