MugenNet: A Novel Combined Convolution Neural Network and Transformer Network with its Application for Colonic Polyp Image Segmentation

Biomedical image segmentation is a very important part in disease diagnosis. The term "colonic polyps" refers to polypoid lesions that occur on the surface of the colonic mucosa within the intestinal lumen. In clinical practice, early detection of polyps is conducted through colonoscopy examinations and biomedical image processing. Therefore, the accurate polyp image segmentation is of great significance in colonoscopy examinations. Convolutional Neural Network (CNN) is a common automatic segmentation method, but its main disadvantage is the long training time. Transformer utilizes a self-attention mechanism, which essentially assigns different importance weights to each piece of information, thus achieving high computational efficiency during segmentation. However, a potential drawback is the risk of information loss. In the study reported in this paper, based on the well-known hybridization principle, we proposed a method to combine CNN and Transformer to retain the strengths of both, and we applied this method to build a system called MugenNet for colonic polyp image segmentation. We conducted a comprehensive experiment to compare MugenNet with other CNN models on five publicly available datasets. The ablation experiment on MugentNet was conducted as well. The experimental results show that MugenNet achieves significantly higher processing speed and accuracy compared with CNN alone. The generalized implication with our work is a method to optimally combine two complimentary methods of machine learning.

PyroTrack: Belief-Based Deep Reinforcement Learning Path Planning for Aerial Wildfire Monitoring in Partially Observable Environments

Motivated by agility, 3D mobility, and low-risk operation compared to human-operated management systems of autonomous unmanned aerial vehicles (UAVs), this work studies UAV-based active wildfire monitoring where a UAV detects fire incidents in remote areas and tracks the fire frontline. A UAV path planning solution is proposed considering realistic wildfire management missions, where a single low-altitude drone with limited power and flight time is available. Noting the limited field of view of commercial low-altitude UAVs, the problem formulates as a partially observable Markov decision process (POMDP), in which wildfire progression outside the field of view causes inaccurate state representation that prevents the UAV from finding the optimal path to track the fire front in limited time. Common deep reinforcement learning (DRL)-based trajectory planning solutions require diverse drone-recorded wildfire data to generalize pre-trained models to real-time systems, which is not currently available at a diverse and standard scale. To narrow down the gap caused by partial observability in the space of possible policies, a belief-based state representation with broad, extensive simulated data is proposed where the beliefs (i.e., ignition probabilities of different grid areas) are updated using a Bayesian framework for the cells within the field of view. The performance of the proposed solution in terms of the ratio of detected fire cells and monitored ignited area (MIA) is evaluated in a complex fire scenario with multiple rapidly growing fire batches, indicating that the belief state representation outperforms the observation state representation both in fire coverage and the distance to fire frontline.

LLM4Decompile: Decompiling Binary Code with Large Language Models

Decompilation aims to restore compiled code to human-readable source code, but struggles with details like names and structure. Large language models (LLMs) show promise for programming tasks, motivating their application to decompilation. However, there does not exist any open-source LLM for decompilation. Moreover, existing decompilation evaluation systems mainly consider token-level accuracy and largely ignore code executability, which is the most important feature of any program. Therefore, we release the first open-access decompilation LLMs ranging from 1B to 33B pre-trained on 4 billion tokens of C source code and the corresponding assembly code. The open-source LLMs can serve as baselines for further development in the field. To ensure practical program evaluation, we introduce Decompile-Eval, the first dataset that considers re-compilability and re-executability for decompilation. The benchmark emphasizes the importance of evaluating the decompilation model from the perspective of program semantics. Experiments indicate that our LLM4Decompile has demonstrated the capability to accurately decompile 21% of the assembly code, which achieves a 50% improvement over GPT-4. Our code, dataset, and models are released at https://github.com/albertan017/LLM4Decompile

$1$-Bit SubTHz RIS with Planar Tightly Coupled Dipoles: Beam Shaping and Prototypes

In this paper, a proof-of-concept study of a $1$-bit wideband reconfigurable intelligent surface (RIS) comprising planar tightly coupled dipoles (PTCD) is presented. The developed RIS operates at subTHz frequencies and a $3$-dB gain bandwidth of $27.4\%$ with the center frequency at $102$ GHz is shown to be obtainable via full-wave electromagnetic simulations. The binary phase shift offered by each RIS unit element is enabled by changing the polarization of the reflected wave by $180^\circ$. The proposed PTCD-based RIS has a planar configuration with one dielectric layer bonded to a ground plane, and hence, it can be fabricated by using cost-effective printed circuit board (PCB) technology. We analytically calculate the response of the entire designed RIS and showcase that a good agreement between that result and equivalent full-wave simulations is obtained. To efficiently compute the $1$-bit RIS response for different pointing directions, thus, designing a directive beam codebook, we devise a fast approximate beamforming optimization approach, which is compared with time-consuming full-wave simulations. Finally, to prove our concept, we present several passive prototypes with frozen beams for the proposed $1$-bit wideband RIS.

Low-rank Adaptation of Large Language Model Rescoring for Parameter-Efficient Speech Recognition

We propose a neural language modeling system based on low-rank adaptation (LoRA) for speech recognition output rescoring. Although pretrained language models (LMs) like BERT have shown superior performance in second-pass rescoring, the high computational cost of scaling up the pretraining stage and adapting the pretrained models to specific domains limit their practical use in rescoring. Here we present a method based on low-rank decomposition to train a rescoring BERT model and adapt it to new domains using only a fraction (0.08%) of the pretrained parameters. These inserted matrices are optimized through a discriminative training objective along with a correlation-based regularization loss. The proposed low-rank adaptation Rescore-BERT (LoRB) architecture is evaluated on LibriSpeech and internal datasets with decreased training times by factors between 5.4 and 3.6.

Rhythm-controllable Attention with High Robustness for Long Sentence Speech Synthesis

Regressive Text-to-Speech (TTS) system utilizes attention mechanism to generate alignment between text and acoustic feature sequence. Alignment determines synthesis robustness (e.g, the occurence of skipping, repeating, and collapse) and rhythm via duration control. However, current attention algorithms used in speech synthesis cannot control rhythm using external duration information to generate natural speech while ensuring robustness. In this study, we propose Rhythm-controllable Attention (RC-Attention) based on Tracotron2, which improves robustness and naturalness simultaneously. Proposed attention adopts a trainable scalar learned from four kinds of information to achieve rhythm control, which makes rhythm control more robust and natural, even when synthesized sentences are extremely longer than training corpus. We use word errors counting and AB preference test to measure robustness of proposed method and naturalness of synthesized speech, respectively. Results shows that RC-Attention has the lowest word error rate of nearly 0.6%, compared with 11.8% for baseline system. Moreover, nearly 60% subjects prefer to the speech synthesized with RC-Attention to that with Forward Attention, because the former has more natural rhythm.

PROCTER: PROnunciation-aware ConTextual adaptER for personalized speech recognition in neural transducers

End-to-End (E2E) automatic speech recognition (ASR) systems used in voice assistants often have difficulties recognizing infrequent words personalized to the user, such as names and places. Rare words often have non-trivial pronunciations, and in such cases, human knowledge in the form of a pronunciation lexicon can be useful. We propose a PROnunCiation-aware conTextual adaptER (PROCTER) that dynamically injects lexicon knowledge into an RNN-T model by adding a phonemic embedding along with a textual embedding. The experimental results show that the proposed PROCTER architecture outperforms the baseline RNN-T model by improving the word error rate (WER) by 44% and 57% when measured on personalized entities and personalized rare entities, respectively, while increasing the model size (number of trainable parameters) by only 1%. Furthermore, when evaluated in a zero-shot setting to recognize personalized device names, we observe 7% WER improvement with PROCTER, as compared to only 1% WER improvement with text-only contextual attention

AutoAttention: Automatic Field Pair Selection for Attention in User Behavior Modeling

In Click-through rate (CTR) prediction models, a user's interest is usually represented as a fixed-length vector based on her history behaviors. Recently, several methods are proposed to learn an attentive weight for each user behavior and conduct weighted sum pooling. However, these methods only manually select several fields from the target item side as the query to interact with the behaviors, neglecting the other target item fields, as well as user and context fields. Directly including all these fields in the attention may introduce noise and deteriorate the performance. In this paper, we propose a novel model named AutoAttention, which includes all item/user/context side fields as the query, and assigns a learnable weight for each field pair between behavior fields and query fields. Pruning on these field pairs via these learnable weights lead to automatic field pair selection, so as to identify and remove noisy field pairs. Though including more fields, the computation cost of AutoAttention is still low due to using a simple attention function and field pair selection. Extensive experiments on the public dataset and Tencent's production dataset demonstrate the effectiveness of the proposed approach.

SLAM-TKA: Real-time Intra-operative Measurement of Tibial Resection Plane in Conventional Total Knee Arthroplasty

Total knee arthroplasty (TKA) is a common orthopaedic surgery to replace a damaged knee joint with artificial implants. The inaccuracy of achieving the planned implant position can result in the risk of implant component aseptic loosening, wear out, and even a joint revision, and those failures most of the time occur on the tibial side in the conventional jig-based TKA (CON-TKA). This study aims to precisely evaluate the accuracy of the proximal tibial resection plane intra-operatively in real-time such that the evaluation processing changes very little on the CON-TKA operative procedure. Two X-ray radiographs captured during the proximal tibial resection phase together with a pre-operative patient-specific tibia 3D mesh model segmented from computed tomography (CT) scans and a trocar pin 3D mesh model are used in the proposed simultaneous localisation and mapping (SLAM) system to estimate the proximal tibial resection plane. Validations using both simulation and in-vivo datasets are performed to demonstrate the robustness and the potential clinical value of the proposed algorithm.

Emergency Vehicles Audio Detection and Localization in Autonomous Driving

Emergency vehicles in service have right-of-way over all other vehicles. Hence, all other vehicles are supposed to take proper actions to yield emergency vehicles with active sirens. As this task requires the cooperation between ears and eyes for human drivers, it also needs audio detection as a supplement to vision-based algorithms for fully autonomous driving vehicles. In urban driving scenarios, we need to know both the existence of emergency vehicles and their relative positions to us to decide the proper actions. We present a novel system from collecting the real-world siren data to the deployment of models using only two cost-efficient microphones. We are able to achieve promising performance for each task separately, especially within the crucial 10m to 50m distance range to react (the size of our ego vehicle is around 5m in length and 2m in width). The recall rate to determine the existence of sirens is 99.16% , the median and mean angle absolute error is 9.64{\deg} and 19.18{\deg} respectively, and the median and mean distance absolute error of 9.30m and 10.58m respectively within that range. We also benchmark various machine learning approaches that can determine the siren existence and sound source localization which includes direction and distance simultaneously within 50ms of latency.