The current large language models (LLMs) mainly lacks this capability: autonomous learning capability. Also, we can not prepare all the data/knowledge in the world for LLMs. We propose Learning In Conversation (LIC) to solve the problem to let human teach machine to refresh/expand its data/knowledge automatically by natural language conversation. LIC system is an AI system that contains many deep learning tasks. LIC system provide the natural language interface to let machine to learn new data/knowledge in conversation interacting with human automatically. It is critical, but except for the developers of the AI system, others have no natural language interface to do this education for machine. Based on large language models (LLMs) conversation ability, we train an additional intent recognition model to determine when the AI system need to learn new data/knowledge. We add a module for editing the training dataset of LLMs. We propose the methods to implement our idea and discuss the reasons why we design these methods.

A document by Sachini Karunathilake . Click on the document to view its contents.

A document by Anton Krynkin . Click on the document to view its contents.

This paper describes the historical trajectory of video conferencing systems, spanning from their earlier mechanical and analog origins in the 1920s to the sophisticated IP-based services delivered from the cloud in the 2020s. Each technological age is examined, highlighting the technical and functional aspects that characterized its evolution. Commercial landmarks of each age are presented providing a comprehensive overview of the most prominent offerings at pivotal moments in the timeline. By examining the past and speculating on the future, this paper aims to provide a holistic understanding of the development, current state, and forthcoming trends in video conferencing technology.

 X is a discrete memoryless binary source , Sequence X undergoes lossless transformation to satisfy the requirement â\euroœEach 1 is separated by one or more zeros“, ”Each 0 is separated by one or two 1s” or similar condition , These conditions are the decision conditions for error detection in channel transmission. A new channel coding method is proposed based on a weighted probability model for lossless coding , The encoding rate and encoding and decoding steps of different conversion methods are different , It is proved that the decoding error probability can reach 0 when the code length is long enough. In the simulation experiment of BPSK signal in AWGN channel, At 0.5 bit rate, the proposed method improves 1.1dB over Polar code when FER is 0.001, and 1.4dB over Polar code when FER is 0.0001.Â

The topology of low-voltage distribution networks (LVDNs) is crucial for system analysis, e.g., distributed energy resources (DERs) integration, network hosting capacity analysis, state estimation, and electric vehicle charging management. However, it is frequently unavailable or incomplete. We develop a data-driven topology identification approach for DNs with a high proportion of underground cables. The proposed approach exploits the fact that underground cables usually follow the street pattern, thus relying on open street map (OSM) and smart meter (SM) data. Three stages compose the proposed approach: In the first stage, a hierarchical minimum spanning tree algorithm is proposed to generate the initial topology with an accurate number of sub-branches from the pre-processed OSM data and peak demand. In the second stage, based on the limited SM data, the location of breakpoints in mesh topology caused by circle roads is verified and reconstructed to guarantee the radial structure of LVDNs. Finally, given multiple incomplete SM datasets, three data-driven optimization models based on a state estimation model are constructed to mitigate the error of cable length induced by OSM data. The feasibility of the proposed topology identification approach is verified on three actual LVDNs in the Netherlands and multiple incomplete SM datasets.

Nanosecond pulsed field ablation (nsPFA) is a promising modality for clinical tissue ablation. The performance of the pulse generator (PG) strongly depends on the switching characteristics of the SiC MOSFET. Currently, many key parameters listed in SiC MOSFET datasheets are determined using double-pulse test circuits with inductive loads, whereas loads in nsPFA-applications are predominantly resistive. This study proposes a test scheme for SiC MOSFETs under nsPFA-like operating conditions. Key parameters to be evaluated are pulsed drain current (IDM), on-resistance (Ron), rise/fall time, turn-on/off delay, and minimum pulse width. The results show significant differences between manufacturer datasheet values and experimental data. For example, one switch under test has an Idm value of 40 A in the datasheet but 153 A in experimental test. This study recommends that custom testing of SiC MOSFETs is essential when designing PGs for nsPFA-applications.

The absence of labeled datasets is one of the most significant challenges in machine learning. It requires both time and assets. Generative Adversarial Networks (GANs) have been tremendously successful at generating data. Although GANs can generate data, yet in standard setups, they are unable to produce the corresponding labels and the resulting data often lacks diversity. In this research, I will use conditional GANs to generate a synthetic data set based on Fashion-MNIST and compare the classification performance of models trained on each dataset.Â

In this study, we introduce an innovative off-tracking optimisation-based framework designed for Generalised N-Trailer vehicles. The task of generating feasible references for the segments of the interconnected vehicle has historically been burdensome. This challenge becomes even more pronounced as the number of trailers increases, particularly when the quantity of trailers fluctuates during operations. Our proposed method address this complexity by autonomously generating pose references for both the tractor and each trailer of the vehicle. This is achieved through a parameterisation of the desired path and attitudes to be traversed. Notably, our approach incorporates the system's model into the optimisation process as a constraint. This guarantees that the generated reference remains within the bounds of feasibility, even when the nominal path does not. Several simulated and field experiments with a tractor pulling 2 trailers show the effectiveness of the proposed method.

The incidence angle dependence of synthetic aperture radar (SAR) backscatter from sea ice has been observed to differ between ice types. This is indicative of its complex dielectric and backscattering properties and cannot be analytically determined or easily modelled. Without assumptions this dependence can only be measured from SAR through multiple observations of the same ice. Due to the sea ice drift, doing so at high resolution is time inefficient and to some extent only feasible for land fast ice. Thus, the incidence angle dependence can currently not be easily related to ground measurements or be fully exploited for classification tasks. By reformulating the problem of incidence angle dependence to a domain transfer task and making use of known physical relations, we can train a Generative Adversarial Network (GAN) to predict the incidence angle dependence from a single Sentinel-1 satellite scene patch-wise using no only local data. The network’s predictions are validated for multiyear and first-year ice backscatter using coincident ICESat-2 altimeter measurements. The network suggests a diverse incidence angle dependence in the HV channel, that so far have been observed only scarsely. The results of this study are a novelty in sea ice remote sensing insofar that artificial intelligence manages to solve a task that would be nearly impossible for a human observer.

In this paper, we develop an economic mechanism for the DSO to exploit the flexibility of the LEM and integrate it into the techno-economic framework for coordinated procurement of DF supported by the novel active network technologies, in this case SOPs. The LEM flexibility mechanism independently clears the LEM, while respecting the network constraints and optimizing network operation cost by leveraging it with the social welfare of the LEM participants. The coordinated procurement framework is based on the bi-level programming approach, where the lower level ensures unconditional clearing of the LEM bids and offers, while the upper level minimizes network operational cost, subject to network constraints and DF originating from all three types of flexibility (i.e., network, contracted, and LEM). Both, the mechanism and the framework can be applied by the DSO to ensure secure and economic network operation while facilitating LEM clearing in the independent manner. In the numerical study, the framework has been applied to IEEE 33-bus benchmark system, where we investigate the performance of the coordination method proposed, evaluate the value of the LEM flexibility mechanism to the DSO, and the value of the novel active network technology (i.e., SOP). Using realistic flexibility demand and supply data, the numerical study showed that the optimal utilization of network and LEM flexibilities can lead to the DSO contracted flexibility service to become unnecessary. While the estimated value of coordinated procurement of the LEM and DSO contracted flexibilities accounts for 26.6% of the network operation cost, technology supported procurement of the LEM flexibility only can reduce the original network operation cost almost twice (by 48.9%).

For wideband transceivers operating at millimeter wave and terahertz frequencies, the implementation of conventional digital predistortion for nonlinearity mitigation faces significant challenges due to the limited availability and/or complexity of high-speed digital signal processing. In this paper, a simple received signal model is proposed for wideband system with nonlinearity and other practical impairments, such as transmitter (Tx) and receiver (Rx) I/Q imbalances (IQIs), carrier frequency offset (CFO), and phase noise, to enable low-complexity impairment mitigation. An expanded memory polynomial (EMP) model is firstly proposed to capture Tx IQI and the nonlinearity over the entire transceiver chain. Exploiting the CFO and a novel transmission protocol, a blind Rx IQI estimation is also proposed. The noise enhancement after Rx IQI and CFO compensation is then evaluated as a noise factor related to the mean-square-error of the Rx IQI estimation. As a result, the received signal of the wideband system is finally modelled as an EMP plus additive noises followed by a band-limited noisy receiver filter. Simulation results using a millimeter wave system with 2.5 GHz bandwidth and 73.5 GHz carrier frequency are presented to verify the accuracy of the EMP modelling and validate the theoretical analyses.

Wideband millimeter wave and terahertz systems face severe nonlinearity and other practical impairments such as transmitter and receiver I/Q imbalances (IQIs), carrier frequency offset, and phase noise. Based on a simplified yet effective received signal model, this paper proposes a low-complexity digital post-cancellation (DPC) framework for transmitter IQI and overall system nonlinearity mitigation. The nonlinearity parameters with reduced nonlinearity order are firstly estimated with low-complexity using a novel transmission protocol incorporating both frame rotation and preamble power scaling. Through widely linear system equalization and interference cancellation, the signal distortion caused by frequency-dependent IQI and nonlinearity are then mitigated with significant performance improvement. The mean-squared-error measurement of the EMP-modelled signals also provides a practical means for the nonlinear system identification and characterization. Both simulation and experiment results obtained from a millimeter wave system with 2.5 GHz bandwidth and 73.5 GHz carrier frequency are presented to verify the theoretical analyses and demonstrate the effectiveness of the DPC technology.

Due to the necessary security for the airport and flight, passengers are required to have strict security check before getting aboard. However, there are frequent complaints of wasting huge amount of time while waiting for the security check. This paper presents a potential solution aimed at optimizing gate setup procedures specifically tailored for Chicago Oâ\euro™Hare International Airport. By referring to queueing theory and performing Monte Carlo simulations, we propose an approach to significantly diminish the average waiting time to a more manageable level. Additionally, our study meticulously examines and identifies the influential factors contributing to this optimization, providing a comprehensive understanding of their impact.

Transformer requires a fixed number of layers and heads which makes them inflexible to the complexity of individual samples and expensive in training and inference. To address this, we propose a sample-based Dynamic Hierarchical Transformer (DHT) model whose layers and heads can be dynamically configured with single data samples via solving contextual bandit problems. To determine the number of layers and heads, we use the Uniform Confidence Bound algorithm while we deploy combinatorial Thompson Sampling in order to select specific head combinations given their number. Different from previous work that focuses on compressing trained networks for inference only, DHT is not only advantageous for adaptively optimizing the underlying network architecture during training but also has a flexible network for efficient inference. To the best of our knowledge, this is the first comprehensive data-driven dynamic transformer without any additional auxiliary neural networks that implement the dynamic system. According to the experiment results, we achieve up to 74\% computational savings for both training and inference with a minimal loss of accuracy.

Love wave (LW) acoustic sensors are promising devices for biochemical detection in the liquid media. However, their application for in-situ biochemical detection especially in the turbid liquid medium is not yet explored. Turbid liquids are complex in nature with both mechanical and electrical characteristics. These characteristics could be reliably estimated with LW acoustic sensor response, where some of the dedicated sensors based on electrochemistry or optical principle show limitations. This paper presents experimental responses of LW acoustic sensor to turbid liquids based on Formazin solutions and its comparison to the spectrophotometer response. Analysis of sensorâ\euro™s electro-acoustic response is used to better characterize the sensor performance with turbid liquids. Further, this paper demonstrates the feasibility of LW acoustic sensor as a multiparameter sensing unit by estimating the influence of electro-mechanical parameters of the turbid liquids on the overall response of the sensor.

Learning in sequential decision making problems can be significantly affected by the choice of reward function. Although crucial for learning success, reward function design remains a complex task, requiring expertise and not always aligned with human preferences. Large Language Models (LLMs) offer a promising avenue for reward design through textual prompts that leverage the prior knowledge and contextual reasoning of LLMs. Despite their potential, LLM responses lack guarantees and their reasoning abilities are poorly understood. To mitigate potential LLM errors, we introduce an alternative approach: learning a new reward function utilizing LLM outputs as auxiliary rewards. This problem is tackled through a bi-level optimization framework, showcasing the method’s proficiency in both optimal reward acquisition and adaptive reward shaping. The proposed approach demonstrates robustness and effectiveness, offering a novel strategy for enhancing reinforcement learning outcomes.     Â

Many companies and a growing number of governments now have guides, frameworks, or principles claiming to govern their use of AI. Seven years ago, virtually none did. This article presents original research on documents produced by 193 countries and the top 100 companies by market capitalization. A key shift occurred in 2016 when the Cambridge Analytica scandal showed the potential harms of misused AI. The widespread use of large language models such as ChatGPT beginning in late 2022 is further increasing calls for governance of the AI space. Analyzing the evolving practice of releasing such documents and the language that they use offers important insights into how norms around AI are spreading and changing â\euro” and where they might go next.