Memorizing All for Implicit Discourse Relation Recognition

Implicit discourse relation recognition is a challenging task due to the absence of the necessary informative clues from explicit connectives. An implicit discourse relation recognizer has to carefully tackle the semantic similarity of sentence pairs and the severe data sparsity issue. In this article, we learn token embeddings to encode the structure of a sentence from a dependency point of view in their representations and use them to initialize a baseline model to make it really strong. Then, we propose a novel memory component to tackle the data sparsity issue by allowing the model to master the entire training set, which helps in achieving further performance improvement. The memory mechanism adequately memorizes information by pairing representations and discourse relations of all training instances, thus filling the slot of the data-hungry issue in the current implicit discourse relation recognizer. The proposed memory component, if attached with any suitable baseline, can help in performance enhancement. The experiments show that our full model with memorizing the entire training data provides excellent results on PDTB and CDTB datasets, outperforming the baselines by a fair margin.

Person Memory Mechanism Underlying Approach and Avoidance Judgments of Social Targets

People display approach and avoidance tendencies toward social targets. Although much research has studied the factors that affect decisions to approach or avoid targets, less work has investigated whether cognitive factors, such as episodic memory (e.g., details remembered about others from previous encounters) contribute to such judgments. Across two experiments, participants formed positive or negative impressions of targets based on their picture, a trait-implying behavior (Experiments 1 & 2), and their political ideology (conservative or liberal; Experiment 2). Memory and approach/avoidance decisions for targets were then measured. Results showed remembering negative impressions about targets increased avoidance responses, whereas remembering positive impressions increased approach responses. Strikingly, falsely remembering negative impressions for novel social targets (not seen before) also induced avoidance. Results suggest remembering negative information about targets, whether correctly or falsely, strongly influences future social judgments. Overall, these data support an episodic memory mechanism underlying subsequent approach/avoidance judgments, which is a rich area for future research.

Fractal structure and hydration-driven shape memory of duck down in the dry–wet state

Duck down, as a natural keratin material, has been widely used as a filling material. The multilevel bifurcation structure of down has been observed and characterized through scanning electron microscopy. The structure is a complex fractal structure composed of four-level self-similar structures including five units, that is, the calamus, main barb, barb, barbule, and node or prong. The differential friction effect of the dynamic friction coefficients of the barb was reduced from 0.4 (dry state) to 0.23 (wet state), namely a decrease of 42.5%. The friction locking effect decreases due to the swelling of the fiber diameter. The down is zero gravity in water, and under the action of vibration and internal stress, down that has been subjected to friction or heat setting treatment can quickly return to its original shape in water. This shape memory mechanism was further confirmed, in which down after heat setting can restore its shape to the natural state by shaking it quickly and vigorously. This research provides inspiration to investigate more complicated functions of natural materials and encourages the creation of very intelligent synthetic polymers.

„Wolałbym pamiętać”. Relacja rzeczy, pamięci i tożsamości w wybranych opowiadaniach Pawła Sołtysa

The article is an attempt to establish the relationship between things, memory, and identity building in Mikrotyki by Paweł Sołtys. Three short stories (Szlugi, Listy, and Patelnia) in which the items constitute the main plotline were selected for analysis. Detailed analysis of the texts showed that all things are strongly connected with the category of nostalgia and the work of memory, thus becoming the building blocks of the writer’s identity, but each of them has a different meaning in this process. Cigarettes are primarily an impulse that activates the memory mechanism and refers to an experience shared by a certain generation. Letters, as nostalgic things,are a memento of the past. An abandoned frying pan becomes a contribution to reflection on the relationship between people and things. Each of the described items forces the author to interpret the past in a specific way, and thus also the present.

Using a new HSPC senescence model in vitro to explore the mechanism of cellular memory in aging HSPCs

Background Age-associated changes attenuate human blood system functionality through the aging of hematopoietic stem and progenitor cells (HSPCs), manifested in human populations an increase in myeloproliferative disease and even leukemia; therefore, study on HSPC senescence bears great significance to treat hematopoietic-associated disease. Furthermore, the mechanism of HSPC aging is lacking, especially the cellular memory mechanism. Here, we not only reported a new HSPC senescence model in vitro, but also propose and verify the cellular memory mechanism of HSPC aging of the Polycomb/Trithorax system. Methods HSPCs (Lin−c-kit+ cells) were isolated and purified by magnetic cell sorting (MACS). The proportions and cell cycle distribution of cells were determined by flow cytometry; senescence-related β-galactosidase assay, transmission electron microscope (TEM), and colony-forming unit (CFU)-mix assay were detected for identification of the old HSPC model. Proteomic tests and RNA-seq were applied to analyze differential pathways and genes in the model cells. qPCR, Western blot (WB), and chromatin immunoprecipitation PCR (CHIP-PCR) were used to detect the gene expression of cell memory-related proteins. Knockdown of cell memory-related key genes was performed with shRNA interference. Results In the model old HSPCs, β-gal activity, cell cycle, colony-forming ability, aging-related cell morphology, and metabolic pathway were significantly changed compared to the young HSPCs. Furthermore, we found the model HSPCs have more obvious aging manifestations than those of natural mice, and IL3 is the major factor contributing to HSPC aging in the model. We also observed dramatic changes in the expression level of PRC/TrxG complexes. After further exploring the downstream molecules of PRC/TrxG complexes, we found that Uhrf1 and TopII played critical roles in HSPC aging based on the HSPC senescence model. Conclusions These findings proposed a new HSPC senescence model in vitro which we forecasted could be used to preliminary screen the drugs of the HSPC aging-related hemopathy and suggested cellular memory mechanism of HSPC aging.

Transcription Control of Liver Development

During liver organogenesis, cellular transcriptional profiles are constantly reshaped by the action of hepatic transcriptional regulators, including FoxA1-3, GATA4/6, HNF1α/β, HNF4α, HNF6, OC-2, C/EBPα/β, Hex, and Prox1. These factors are crucial for the activation of hepatic genes that, in the context of compact chromatin, cannot access their targets. The initial opening of highly condensed chromatin is executed by a special class of transcription factors known as pioneer factors. They bind and destabilize highly condensed chromatin and facilitate access to other “non-pioneer” factors. The association of target genes with pioneer and non-pioneer transcription factors takes place long before gene activation. In this way, the underlying gene regulatory regions are marked for future activation. The process is called “bookmarking”, which confers transcriptional competence on target genes. Developmental bookmarking is accompanied by a dynamic maturation process, which prepares the genomic loci for stable and efficient transcription. Stable hepatic expression profiles are maintained during development and adulthood by the constant availability of the main regulators. This is achieved by a self-sustaining regulatory network that is established by complex cross-regulatory interactions between the major regulators. This network gradually grows during liver development and provides an epigenetic memory mechanism for safeguarding the optimal expression of the regulators.

Controlling reversible cell differentiation for labor division in microbial consortia

We address the problem of regulating and keeping at a desired balance the relative numbers between cells exhibiting a different phenotype within a monostrain microbial consortium. We propose a strategy based on the use of external control inputs, assuming each cell in the community is endowed with a reversible, bistable memory mechanism. Specifically, we provide a general analytical framework to guide the design of external feedback control strategies aimed at balancing the ratio between cells whose memory is stabilized at either one of two equilibria associated to different cell phenotypes. We demonstrate the stability and robustness properties of the control laws proposed and validate them in silico implementing the memory element via a genetic toggle-switch. The proposed control framework may be used to allow long term coexistence of different populations, with both industrial and biotechnological applications. Examples include consortia where each population produces a compound of interest or where one population supports the growth of the other which has the role of producing a desired molecule. As a representative example we consider the realistic agent-based implementation of our control strategy to enable cooperative bioproduction in microbial consortia.

Learn from Concepts: Towards the Purified Memory for Few-shot Learning

Human beings have a great generalization ability to recognize a novel category by only seeing a few number of samples. This is because humans possess the ability to learn from the concepts that already exist in our minds. However, many existing few-shot approaches fail in addressing such a fundamental problem, {\it i.e.,} how to utilize the knowledge learned in the past to improve the prediction for the new task. In this paper, we present a novel purified memory mechanism that simulates the recognition process of human beings. This new memory updating scheme enables the model to purify the information from semantic labels and progressively learn consistent, stable, and expressive concepts when episodes are trained one by one. On its basis, a Graph Augmentation Module (GAM) is introduced to aggregate these concepts and knowledge learned from new tasks via a graph neural network, making the prediction more accurate. Generally, our approach is model-agnostic and computing efficient with negligible memory cost. Extensive experiments performed on several benchmarks demonstrate the proposed method can consistently outperform a vast number of state-of-the-art few-shot learning methods.