skip to main content


Title: Blocking Influence at Collective Level with Hard Constraints (Student Abstract)
Influence blocking maximization (IBM) is crucial in many critical real-world problems such as rumors prevention and epidemic containment. The existing work suffers from: (1) concentrating on uniform costs at the individual level, (2) mostly utilizing greedy approaches to approximate optimization, (3) lacking a proper graph representation for influence estimates. To address these issues, this research introduces a neural network model dubbed Neural Influence Blocking (\algo) for improved approximation and enhanced influence blocking effectiveness. The code is available at https://github.com/oates9895/NIB.  more » « less
Award ID(s):
2153369
NSF-PAR ID:
10408644
Author(s) / Creator(s):
; ; ; ; ; ; ;
Date Published:
Journal Name:
Proceedings of the AAAI Conference on Artificial Intelligence
Volume:
36
Issue:
11
ISSN:
2159-5399
Page Range / eLocation ID:
13115 to 13116
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Abstract

    Atmospheric blocking events are persistent quasi‐stationary geopotential height anomalies that divert the jet stream from its climatological path in the mid‐ to high‐latitudes. Previous studies have found that different phases of the El Niño–Southern Oscillation (ENSO) influence the characteristics of blocking, but none have considered the spatial diversity of El Niño. In this study, we examine Northern Hemisphere blocking events with respect to the “Central Pacific” (CP) and “Eastern Pacific” (EP) flavors of El Niño in 83 years of ERA5 reanalysis. The two El Niño flavors have dissimilar patterns of forcing on atmospheric circulation that impact the strength and placement of the upper‐level jet stream, thus affecting blocking event frequency and duration. Significant contrasts in blocking characteristics between CP and EP years are disregarded when a single ENSO index is used, and we emphasize that El Niño flavors should be considered in future investigations of blocking and ENSO‐related variability.

     
    more » « less
  2. The availability of massive data and computing allowing for effective data driven neural approaches is having a major impact on AI and IR research, but these models have a basic problem with efficiency. Current neural ranking models are implemented as multistage rankers: for efficiency reasons, the neural model only re-ranks the top ranked documents retrieved by a first-stage efficient ranker in response to a given query. Neural ranking models learn dense representations causing essentially every query term to match every document term, making it highly inefficient or intractable to rank the whole collection. The reliance on a first stage ranker creates a dual problem: First, the interaction and combination effects are not well understood. Second, the first stage ranker serves as a "gate-keeper" or filter effectively blocking the potential of neural models to uncover new relevant documents. In this work, we propose a standalone neural ranking model SNRM by introducing a sparsity property to learn a latent sparse representation for each query and document. This representation captures the semantic relationship between the query and documents, but is also {sparse} enough to enable constructing an inverted index for the whole collection. We parameterize the sparsity of the model to yield a retrieval model as efficient as conventional term based models. Our model gains in efficiency without loss of effectiveness: it not only outperforms the existing term matching baselines, but also performs similar to the recent re-ranking based neural models with dense representations. More generally, our results demonstrate the importance of sparsity in neural model learning and show that dense representations can be pruned effectively, giving new insights about essential semantic features and their distributions. 
    more » « less
  3. Neural network flexibility includes changes in neuronal participation between networks, such as the switching of neurons between single- and dual-network activity. We previously identified a neuron that is recruited to burst in time with an additional network via modulation of its intrinsic membrane properties, instead of being recruited synaptically into the second network. However, the modulated intrinsic properties were not determined. Here, we use small networks in the Jonah crab ( Cancer borealis) stomatogastric nervous system (STNS) to examine modulation of intrinsic properties underlying neuropeptide (Gly 1 -SIFamide)-elicited neuronal switching. The lateral posterior gastric neuron (LPG) switches from exclusive participation in the fast pyloric (∼1 Hz) network, due to electrical coupling, to dual-network activity that includes periodic escapes from the fast rhythm via intrinsically generated oscillations at the slower gastric mill network frequency (∼0.1 Hz). We isolated LPG from both networks by pharmacology and hyperpolarizing current injection. Gly 1 -SIFamide increased LPG intrinsic excitability and rebound from inhibition and decreased spike frequency adaptation, which can all contribute to intrinsic bursting. Using ion substitution and channel blockers, we found that a hyperpolarization-activated current, a persistent sodium current, and calcium or calcium-related current(s) appear to be primary contributors to Gly 1 -SIFamide-elicited LPG intrinsic bursting. However, this intrinsic bursting was more sensitive to blocking currents when LPG received rhythmic electrical coupling input from the fast network than in the isolated condition. Overall, a switch from single- to dual-network activity can involve modulation of multiple intrinsic properties, while synaptic input from a second network can shape the contributions of these properties. NEW & NOTEWORTHY Neuropeptide-elicited intrinsic bursting was recently determined to switch a neuron from single- to dual-network participation. Here we identified multiple intrinsic properties modulated in the dual-network state and candidate ion channels underlying the intrinsic bursting. Bursting at the second network frequency was more sensitive to blocking currents in the dual-network state than when neurons were synaptically isolated from their home network. Thus, synaptic input can shape the contributions of modulated intrinsic properties underlying dual-network activity. 
    more » « less
  4. The acoustic environment an animal experiences early in life shapes the structure and function of its auditory system. This process of experience-dependent development is thought to be primarily orchestrated by potentiation and depression of synapses, but plasticity of intrinsic voltage dynamics may also contribute. Here, we show that in juvenile male and female zebra finches, neurons in a cortical-level auditory area, the caudal mesopallium (CM), can rapidly change their firing dynamics. This plasticity was only observed in birds that were reared in a complex acoustic and social environment, which also caused increased expression of the low-threshold potassium channel Kv1.1 in the plasma membrane and endoplasmic reticulum (ER). Intrinsic plasticity depended on activity, was reversed by blocking low-threshold potassium currents, and was prevented by blocking intracellular calcium signaling. Taken together, these results suggest that Kv1.1 is rapidly mobilized to the plasma membrane by activity-dependent elevation of intracellular calcium. This produces a shift in the excitability and temporal integration of CM neurons that may be permissive for auditory learning in complex acoustic environments during a crucial period for the development of vocal perception and production.

    SIGNIFICANCE STATEMENTNeurons can change not only the strength of their connections to other neurons, but also how they integrate synaptic currents to produce patterns of action potentials. In contrast to synaptic plasticity, the mechanisms and functional roles of intrinisic plasticity remain poorly understood. We found that neurons in the zebra finch auditory cortex can rapidly shift their spiking dynamics within a few minutes in response to intracellular stimulation. This plasticity involves increased conductance of a low-threshold potassium current associated with the Kv1.1 channel, but it only occurs in birds reared in a rich acoustic environment. Thus, auditory experience regulates a mechanism of neural plasticity that allows neurons to rapidly adapt their firing dynamics to stimulation.

     
    more » « less
  5. Hierarchical text classification, which aims to classify text documents into a given hierarchy, is an important task in many real-world applications. Recently, deep neural models are gaining increasing popularity for text classification due to their expressive power and minimum requirement for feature engineering. However, applying deep neural networks for hierarchical text classification remains challenging, because they heavily rely on a large amount of training data and meanwhile cannot easily determine appropriate levels of documents in the hierarchical setting. In this paper, we propose a weakly-supervised neural method for hierarchical text classification. Our method does not require a large amount of training data but requires only easy-to-provide weak supervision signals such as a few class-related documents or keywords. Our method effectively leverages such weak supervision signals to generate pseudo documents for model pre-training, and then performs self-training on real unlabeled data to iteratively refine the model. During the training process, our model features a hierarchical neural structure, which mimics the given hierarchy and is capable of determining the proper levels for documents with a blocking mechanism. Experiments on three datasets from different domains demonstrate the efficacy of our method compared with a comprehensive set of baselines. 
    more » « less