Search

This paper introduces a Spiking Diffusion Policy (SDP) learning method for robotic manipulation by integrating Spiking Neurons and Learnable Channel-wise Membrane Thresholds (LCMT) into the diffusion policy model, thereby enhancing computational efficiency and achieving high performance in evaluated tasks. Specifically, the proposed SDP model employs the U-Net architecture as the backbone for diffusion learning within the Spiking Neural Network (SNN). It strategically places residual connections between the spike convolution operations and the Leaky Integrate-and-Fire (LIF) nodes, thereby preventing disruptions to the spiking states. Additionally, we introduce a temporal encoding block and a temporal decoding block to transform static and dynamic data with timestep $T_S$ into each other, enabling the transmission of data within the SNN in spike format. Furthermore, we propose LCMT to enable the adaptive acquisition of membrane potential thresholds, thereby matching the conditions of varying membrane potentials and firing rates across channels and avoiding the cumbersome process of manually setting and tuning hyperparameters. Evaluating the SDP model on seven distinct tasks with SNN timestep $T_S=4$, we achieve results comparable to those of the ANN counterparts, along with faster convergence speeds than the baseline SNN method. This improvement is accompanied by a reduction of 94.3\% in dynamic energy consumption estimated on 45nm hardware.

Multiple Sclerosis (MS) is a chronic autoimmune and inflammatory neurological disorder characterised by episodes of symptom exacerbation, known as relapses. In this study, we investigate the role of environmental factors in relapse occurrence among MS patients, using data from the H2020 BRAINTEASER project. We employed predictive models, including Random Forest (RF) and Logistic Regression (LR), with varying sets of input features to predict the occurrence of relapses based on clinical and pollutant data collected over a week. The RF yielded the best result, with an AUC-ROC score of 0.713. Environmental variables, such as precipitation, NO2, PM2.5, humidity, and temperature, were found to be relevant to the prediction.

Multi-modal image fusion aims to combine information from different modes to create a single image with comprehensive information and detailed textures. However, fusion models based on convolutional neural networks encounter limitations in capturing global image features due to their focus on local convolution operations. Transformer-based models, while excelling in global feature modeling, confront computational challenges stemming from their quadratic complexity. Recently, the Selective Structured State Space Model has exhibited significant potential for long-range dependency modeling with linear complexity, offering a promising avenue to address the aforementioned dilemma. In this paper, we propose FusionMamba, a novel dynamic feature enhancement method for multimodal image fusion with Mamba. Specifically, we devise an improved efficient Mamba model for image fusion, integrating efficient visual state space model with dynamic convolution and channel attention. This refined model not only upholds the performance of Mamba and global modeling capability but also diminishes channel redundancy while enhancing local enhancement capability. Additionally, we devise a dynamic feature fusion module (DFFM) comprising two dynamic feature enhancement modules (DFEM) and a cross modality fusion mamba module (CMFM). The former serves for dynamic texture enhancement and dynamic difference perception, whereas the latter enhances correlation features between modes and suppresses redundant intermodal information. FusionMamba has yielded state-of-the-art (SOTA) performance across various multimodal medical image fusion tasks (CT-MRI, PET-MRI, SPECT-MRI), infrared and visible image fusion task (IR-VIS) and multimodal biomedical image fusion dataset (GFP-PC), which is proved that our model has generalization ability. The code for FusionMamba is available at https://github.com/millieXie/FusionMamba.

We propose a novel approach to iterated sparse matrix dense matrix multiplication, a fundamental computational kernel in scientific computing and graph neural network training. In cases where matrix sizes exceed the memory of a single compute node, data transfer becomes a bottleneck. An approach based on dense matrix multiplication algorithms leads to suboptimal scalability and fails to exploit the sparsity in the problem. To address these challenges, we propose decomposing the sparse matrix into a small number of highly structured matrices called arrow matrices, which are connected by permutations. Our approach enables communication-avoiding multiplications, achieving a polynomial reduction in communication volume per iteration for matrices corresponding to planar graphs and other minor-excluded families of graphs. Our evaluation demonstrates that our approach outperforms a state-of-the-art method for sparse matrix multiplication on matrices with hundreds of millions of rows, offering near-linear strong and weak scaling.

In this paper, we analyze the performance of a multitask end-to-end transformer model on the task of conversational recommendations, which aim to provide recommendations based on a user's explicit preferences expressed in dialogue. While previous works in this area adopt complex multi-component approaches where the dialogue management and entity recommendation tasks are handled by separate components, we show that a unified transformer model, based on the T5 text-to-text transformer model, can perform competitively in both recommending relevant items and generating conversation dialogue. We fine-tune our model on the ReDIAL conversational movie recommendation dataset, and create additional training tasks derived from MovieLens (such as the prediction of movie attributes and related movies based on an input movie), in a multitask learning setting. Using a series of probe studies, we demonstrate that the learned knowledge in the additional tasks is transferred to the conversational setting, where each task leads to a 9%-52% increase in its related probe score., Comment: 10 pages, 4 tables, 1 figure

The purpose of this paper is to study the fractal phenomena in large data sets and the associated questions of dimension reduction. We examine situations where the classical Principal Component Analysis is not effective in identifying the salient underlying fractal features of the data set. Instead, we employ the discrete energy, a technique borrowed from geometric measure theory, to limit the number of points of a given data set that lie near a $k$-dimensional hyperplane, or, more generally, near a set of a given upper Minkowski dimension. Concrete motivations stemming from naturally arising data sets are described and future directions outlined.

There is growing interest in transitioning electronic components and circuitry from stiff and rigid substrates to more flexible and stretchable platforms, such as thin plastics, textiles, and foams. In parallel, the push for more sustainable, biocompatible, and cost-efficient conductive inks to coat these substrates, has led to the development of formulations with novel nanomaterials. Among these, 2D materials, and particularly graphenes and MXenes, have received intense research interest due to their increasingly facile and scalable production, high electrical conductivity, and compatibility with existing manufacturing techniques. They enable a range of electronic devices, including strain and pressure sensors, supercapacitors, thermoelectric generators, and heaters. These new flexible and stretchable electronic devices developed with 2D material coatings are poised to unlock exciting applications in the wearable, healthcare and Internet of Things sectors. This review has surveyed key data from more than 200 articles published over the last 6 years, to provide a quantitative analysis of recent progress in the field and shade light on future directions and prospects of this technology. We find that despite the different chemical origins of graphenes and MXenes, their shared electrical properties and 2D morphology, guarantee intriguing performance in end applications, leaving plenty of space for shared progress and advancements in the future.

In recent years, more and more investors use technical analysis methods in their own trading. Evaluating the effectiveness of technical analysis has become more feasible due to increasing computing capability and blooming public data, which indie investors can perform stock analysis and backtest their own trading strategy conveniently. The Moving Average Convergence Divergence (MACD) indicator is one of the popular technical indicators that are widely used in different strategies. In order to verify the MACD effectiveness, in this thesis, I use the MACD indicator with traditional parameters (12, 26, 9) to build various trading strategies. Then, I apply these strategies to stocks listed on three indices in the US stock market (i.e., Dow-Jones, Nasdaq, and S&P 500) and evaluate its performance in terms of win rate, profitability, Sharpe ratio, number of trades and maximum drawdown. The backtesting is programmed using Python, covering the period between 01/01/2015 and 28-08-2021. The result shows that the win-rate of the strategy with only the MACD indicator is less than 50%. However, the win-rate is improved for the trading strategies that combine the MACD indicator with other momentum indicators like the Money Flow Index (MFI) and the Relative Strength Index (RSI). Based on this result, I redesign the MACD mathematical formula by taking the trading volume and daily price volatility into consideration to derive a new indicator called VPVMA. The results show that the win-rate and risk-adjust performance of this new trading strategy have been improved significantly. In general, the findings suggest that while all the MACD trading strategies mentioned above can generate positive returns, the performance is not good without using other momentum indicators. Hence, the VPVMA indicator performs better.

As more and more online search queries come from voice, automatic speech recognition becomes a key component to deliver relevant search results. Errors introduced by automatic speech recognition (ASR) lead to irrelevant search results returned to the user, thus causing user dissatisfaction. In this paper, we introduce an approach, Mondegreen, to correct voice queries in text space without depending on audio signals, which may not always be available due to system constraints or privacy or bandwidth (for example, some ASR systems run on-device) considerations. We focus on voice queries transcribed via several proprietary commercial ASR systems. These queries come from users making internet, or online service search queries. We first present an analysis showing how different the language distribution coming from user voice queries is from that in traditional text corpora used to train off-the-shelf ASR systems. We then demonstrate that Mondegreen can achieve significant improvements in increased user interaction by correcting user voice queries in one of the largest search systems in Google. Finally, we see Mondegreen as complementing existing highly-optimized production ASR systems, which may not be frequently retrained and thus lag behind due to vocabulary drifts., Comment: Accepted in KDD 2021

This paper investigates the nature of the development of vortex shedding for two-dimensional unsteady flow of an incompressible fluid at the rear stagnation point., Comment: 11 pages, 8 figures, similarity solution of rear stagnation-point flow

The size of deep neural networks (DNNs) grows rapidly as the complexity of the machine learning algorithm increases. To satisfy the requirement of computation and memory of DNN training, distributed deep learning based on model parallelism has been widely recognized. We propose a new pipeline parallelism training framework, BaPipe, which can automatically explore pipeline parallelism training methods and balanced partition strategies for DNN distributed training. In BaPipe, each accelerator calculates the forward propagation and backward propagation of different parts of networks to implement the intra-batch pipeline parallelism strategy. BaPipe uses a new load balancing automatic exploration strategy that considers the parameters of DNN models and the computation, memory, and communication resources of accelerator clusters. We have trained different DNNs such as VGG-16, ResNet-50, and GNMT on GPU clusters and simulated the performance of different FPGA clusters. Compared with state-of-the-art data parallelism and pipeline parallelism frameworks, BaPipe provides up to 3.2x speedup and 4x memory reduction in various platforms.

Many recent advances in neural information retrieval models, which predict top-K items given a query, learn directly from a large training set of (query, item) pairs. However, they are often insufficient when there are many previously unseen (query, item) combinations, often referred to as the cold start problem. Furthermore, the search system can be biased towards items that are frequently shown to a query previously, also known as the 'rich get richer' (a.k.a. feedback loop) problem. In light of these problems, we observed that most online content platforms have both a search and a recommender system that, while having heterogeneous input spaces, can be connected through their common output item space and a shared semantic representation. In this paper, we propose a new Zero-Shot Heterogeneous Transfer Learning framework that transfers learned knowledge from the recommender system component to improve the search component of a content platform. First, it learns representations of items and their natural-language features by predicting (item, item) correlation graphs derived from the recommender system as an auxiliary task. Then, the learned representations are transferred to solve the target search retrieval task, performing query-to-item prediction without having seen any (query, item) pairs in training. We conduct online and offline experiments on one of the world's largest search and recommender systems from Google, and present the results and lessons learned. We demonstrate that the proposed approach can achieve high performance on offline search retrieval tasks, and more importantly, achieved significant improvements on relevance and user interactions over the highly-optimized production system in online experiments., Comment: Accepted at CIKM 2020

This paper explores the data cleaning challenges that arise in using WiFi connectivity data to locate users to semantic indoor locations such as buildings, regions, rooms. WiFi connectivity data consists of sporadic connections between devices and nearby WiFi access points (APs), each of which may cover a relatively large area within a building. Our system, entitled semantic LOCATion cleanER (LOCATER), postulates semantic localization as a series of data cleaning tasks - first, it treats the problem of determining the AP to which a device is connected between any two of its connection events as a missing value detection and repair problem. It then associates the device with the semantic subregion (e.g., a conference room in the region) by postulating it as a location disambiguation problem. LOCATER uses a bootstrapping semi-supervised learning method for coarse localization and a probabilistic method to achieve finer localization. The paper shows that LOCATER can achieve significantly high accuracy at both the coarse and fine levels.

Users issue queries to Search Engines, and try to find the desired information in the results produced. They repeat this process if their information need is not met at the first place. It is crucial to identify the important words in a query that depict the actual information need of the user and will determine the course of a search session. To this end, we propose a sequence-to-sequence based neural architecture that leverages the set of past queries issued by users, and results that were explored by them. Firstly, we employ our model for predicting the words in the current query that are important and would be retained in the next query. Additionally, as a downstream application of our model, we evaluate it on the widely popular task of next query suggestion. We show that our intuitive strategy of capturing information need can yield superior performance at these tasks on two large real-world search log datasets.

Associated to any finite simple graph $\Gamma$ is the chromatic polynomial $P_\Gamma(q)$ whose complex zeroes are called the chromatic zeros of $\Gamma$. A hierarchical lattice is a sequence of finite simple graphs $\{\Gamma_n\}_{n=0}^\infty$ built recursively using a substitution rule expressed in terms of a generating graph. For each $n$, let $\mu_n$ denote the probability measure that assigns a Dirac measure to each chromatic zero of $\Gamma_n$. Under a mild hypothesis on the generating graph, we prove that the sequence $\mu_n$ converges to some measure $\mu$ as $n$ tends to infinity. We call $\mu$ the limiting measure of chromatic zeros associated to $\{\Gamma_n\}_{n=0}^\infty$. In the case of the Diamond Hierarchical Lattice we prove that the support of $\mu$ has Hausdorff dimension two. The main techniques used come from holomorphic dynamics and more specifically the theories of activity/bifurcation currents and arithmetic dynamics. We prove a new equidistribution theorem that can be used to relate the chromatic zeros of a hierarchical lattice to the activity current of a particular marked point. We expect that this equidistribution theorem will have several other applications., Comment: To appear in Annales de l'Institut Henri Poincar\'e D. We have added considerably more background on activity currents and especially on the Dujardin-Favre classification of the passive locus. Exposition in the proof of the main theorem was improved. Comments welcome!

This paper is devoted to an in-depth study of the limiting measure of Lee--Yang zeroes for the Ising Model on the Cayley Tree. We build on previous works of M\"uller-Hartmann-Zittartz (1974 and 1977), Barata--Marchetti (1997), and Barata--Goldbaum (2001), to determine the support of the limiting measure, prove that the limiting measure is not absolutely continuous with respect to Lebesgue measure, and determine the pointwise dimension of the measure at Lebesgue a.e. point on the unit circle and every temperature. The latter is related to the critical exponents for the phase transitions in the model as one crosses the unit circle at Lebesgue a.e. point, providing a global version of the "phase transition of continuous order" discovered by M\"uller-Hartmann-Zittartz. The key techniques are from dynamical systems because there is an explicit formula for the Lee-Yang zeros of the finite Cayley Tree of level $n$ in terms of the $n$-th iterate of an expanding Blaschke Product. A subtlety arises because the conjugacies between Blaschke Products at different parameter values are not absolutely continuous., Comment: Final Version accepted to appear in Communications in Mathematical Physics (some easy proofs that were cut from the published version are still presented here)