The advent of Graph Neural Networks (GNNs) has revolutionized the field of machine learning, offering a novel paradigm for learning on graph-structured data. Unlike traditional neural networks, GNNs are capable of capturing complex relationships and dependencies inherent in graph data, making them particularly suited for a wide range of applications including social network analysis, molecular chemistry, and network security. The impact of GNNs in these domains is profound, enabling more accurate models and predictions, and thereby contributing significantly to advancements in these fields. GNNs, with their unique structure and operation, present new computational challenges compared to conventional neural networks. This requires comprehensive benchmarking and a thorough characterization of GNNs to obtain insight into their computational requirements and to identify potential performance bottlenecks. In this thesis, we aim to develop a better understanding of how GNNs interact with the underlying hardware and will leverage this knowledge as we design specialized accelerators and develop new optimizations, leading to more efficient and faster GNN computations. Synthesizing these insights and optimizations, we design a state-of-the-art hardware accelerator capable of efficiently handling various GNN workloads. Our accelerator architecture is built on our characterization of GNN computational demands, providing clear motivation for our approach. Furthermore, we extend our exploration to emerging GNN workloads in the domain of graph neural networks. This exploration into novel models underlines our comprehensive approach, as we strive to enable accelerators that are not just performant, but also versatile, able to adapt to the evolving landscape of graph computing.

Deep Learning models have achieved remarkable performance in tasks such as image classification or generation, often surpassing human accuracy. However, they can struggle to learn new tasks and update their knowledge without access to previous data, leading to a significant loss of accuracy known as Catastrophic Forgetting (CF). This phenomenon was first observed by McCloskey and Cohen in 1989 and remains an active research topic. Incremental learning without forgetting is widely recognized as a crucial aspect in building better AI systems, as it allows models to adapt to new tasks without losing the ability to perform previously learned ones. This article surveys recent studies that tackle CF in modern Deep Learning models that use gradient descent as their learning algorithm. Although several solutions have been proposed, a definitive solution or consensus on assessing CF is yet to be established. The article provides a comprehensive review of recent solutions, proposes a taxonomy to organize them, and identifies research gaps in this area.

High demand for painters is required nowadays and foreseen in the near future for both developed and developing countries. To satisfy such demand, this paper presents the detailed computer aided design (CAD) model of a fully functional wall painting robot for interior finishes. The RoboPainter is capable of performing full scale wall-ceil painting in addition to decorative wall drawings. The 8 degrees of freedom (DOF) mobile robot structure consists of a 6DOF spray painting arm mounted on a 2DOF differentially driven mobile base. The design presented endows several achievements in terms of total robot mass and painting rate as compared to existing literature. Detailed dynamic model parameters are presented to allow for further enhancement in terms of robot motion control.

Transformer-based large language models (e.g., BERT and GPT) achieve great success, and fine-tuning, which tunes a pre-trained model on a task-specific dataset, is the standard practice to utilize these models for downstream tasks. However, Transformer fine-tuning has long running time and high memory consumption due to the large size of the models. We propose the SPT system to fine-tune Transformer-based models efficiently by introducing sparsity. We observe that the memory consumption of Transformer mainly comes from storing attention weights for multi-head attention (MHA), and the majority of running time is spent on feed-forward network (FFN). Thus, we design the sparse MHA module, which computes and stores only large attention weights to reduce memory consumption, and the routed FFN module, which dynamically activates a subset of model parameters for each token to reduce computation cost. We implement SPT on PyTorch and customize CUDA kernels to run sparse MHA and routed FFN efficiently. Specifically, we use product quantization to identify the large attention weights and compute attention via sparse matrix multiplication for sparse MHA. For routed FFN, we batch the tokens according to their activated model parameters for efficient computation. We conduct extensive experiments to evaluate SPT on various model configurations. The results show that SPT consistently outperforms well-optimized baselines, reducing the peak memory consumption by up to 50% and accelerating fine-tuning by up to 2.2x.

Random Forest is a machine learning method that offers many advantages, including the ability to easily measure variable importance. Class balancing technique is a well-known solution to deal with class imbalance problem. However, it has not been actively studied on RF variable importance. In this paper, we study the effect of class balancing on RF variable importance. Our simulation results show that over-sampling is effective in correctly measuring variable importance in class imbalanced situations with small sample size, while under-sampling fails to differentiate important and non-informative variables. We then propose a variable selection algorithm that utilizes RF variable importance and its confidence interval. Through an experimental study using many real and artificial datasets, we demonstrate that our proposed algorithm efficiently selects an optimal feature set, leading to improved prediction performance in class imbalance problem.

Vegetation structure mapping is critical for understanding the global carbon cycle and monitoring nature-based approaches to climate adaptation and mitigation. Repeated measurements of these data allow for the observation of deforestation or degradation of existing forests, natural forest regeneration, and the implementation of sustainable agricultural practices like agroforestry. Assessments of tree canopy height and crown projected area at a high spatial resolution are also important for monitoring carbon fluxes and assessing tree-based land uses, since forest structures can be highly spatially heterogeneous, especially in agroforestry systems. Very high resolution satellite imagery (less than one meter (1m) Ground Sample Distance) makes it possible to extract information at the tree level while allowing monitoring at a very large scale. This paper presents the first high-resolution canopy height map concurrently produced for multiple sub-national jurisdictions. Specifically, we produce very high resolution canopy height maps for the states of California and Sao Paulo, a significant improvement in resolution over the ten meter (10m) resolution of previous Sentinel / GEDI based worldwide maps of canopy height. The maps are generated by the extraction of features from a self-supervised model trained on Maxar imagery from 2017 to 2020, and the training of a dense prediction decoder against aerial lidar maps. We also introduce a post-processing step using a convolutional network trained on GEDI observations. We evaluate the proposed maps with set-aside validation lidar data as well as by comparing with other remotely sensed maps and field-collected data, and find our model produces an average Mean Absolute Error (MAE) of 2.8 meters and Mean Error (ME) of 0.6 meters.

The computational cost of transformer models makes them inefficient in low-latency or low-power applications. While techniques such as quantization or linear attention can reduce the computational load, they may incur a reduction in accuracy. In addition, globally reducing the cost for all inputs may be sub-optimal. We observe that for each layer, the full width of the layer may be needed only for a small subset of tokens inside a batch and that the "effective" width needed to process a token can vary from layer to layer. Motivated by this observation, we introduce the Adaptive Computation Module (ACM), a generic module that dynamically adapts its computational load to match the estimated difficulty of the input on a per-token basis. An ACM consists of a sequence of learners that progressively refine the output of their preceding counterparts. An additional gating mechanism determines the optimal number of learners to execute for each token. We also describe a distillation technique to replace any pre-trained model with an "ACMized" variant. The distillation phase is designed to be highly parallelizable across layers while being simple to plug-and-play into existing networks. Our evaluation of transformer models in computer vision and speech recognition demonstrates that substituting layers with ACMs significantly reduces inference costs without degrading the downstream accuracy for a wide interval of user-defined budgets.

Ultrasound (US) imaging is a vital adjunct to mammography in breast cancer screening and diagnosis, but its reliance on hand-held transducers often lacks repeatability and heavily depends on sonographers' skills. Integrating US systems from different vendors further complicates clinical standards and workflows. This research introduces a co-robotic US platform for repeatable, accurate, and vendor-independent breast US image acquisition. The platform can autonomously perform 3D volume scans or swiftly acquire real-time 2D images of suspicious lesions. Utilizing a Universal Robot UR5 with an RGB camera, a force sensor, and an L7-4 linear array transducer, the system achieves autonomous navigation, motion control, and image acquisition. The calibrations, including camera-mammogram, robot-camera, and robot-US, were rigorously conducted and validated. Governed by a PID force control, the robot-held transducer maintains a constant contact force with the compression plate during the scan for safety and patient comfort. The framework was validated on a lesion-mimicking phantom. Our results indicate that the developed co-robotic US platform promises to enhance the precision and repeatability of breast cancer screening and diagnosis. Additionally, the platform offers straightforward integration into most mammographic devices to ensure vendor-independence.

The target of Electronic Health Record (EHR) coding is to find the diagnostic codes according to the EHRs. In previous research, researchers have preferred to do multi-classification on the EHR coding task; most of them encode the EHR first and then process it to get the probability of each code based on the EHR representation. However, the question of complicating diseases is neglected among all these methods. In this paper, we propose a novel EHR coding framework, which is the first attempt at detecting complicating diseases, called Copy Recurrent Neural Network Structure Network (CRNNet). This method refers to the idea of adversarial learning; a Path Generator and a Path Discriminator are designed to more efficiently finish the task of EHR coding. We propose a copy module to detect complicating diseases; by the proposed copy module and the adversarial learning strategy, we identify complicating diseases efficiently. Extensive experiments show that our method achieves a 57.30\% ratio of complicating diseases in predictions, demonstrating the effectiveness of our proposed model. According to the ablation study, the proposed copy mechanism plays a crucial role in detecting complicating diseases.

Objective: Function is increasingly recognized as an important indicator of whole-person health. This study evaluates the ability of publicly available large language models (LLMs) to accurately identify the presence of functioning information from clinical notes. We explore various strategies to improve the performance on this task. Materials and Methods: We collect a balanced binary classification dataset of 1000 sentences from the Mobility NER dataset, which was curated from n2c2 clinical notes. For evaluation, we construct zero-shot and few-shot prompts to query the LLMs whether a given sentence contains mobility functioning information. Two sampling techniques, random sampling and k-nearest neighbor (kNN)-based sampling, are used to select the few-shot examples. Furthermore, we apply a parameter-efficient prompt-based fine-tuning method to the LLMs and evaluate their performance under various training settings. Results: Flan-T5-xxl outperforms all other models in both zero-shot and few-shot settings, achieving a F1 score of 0.865 with a single demonstrative example selected by kNN sampling. In prompt-based fine-tuning experiments, this foundation model also demonstrates superior performance across all low-resource settings, particularly achieving an impressive F1 score of 0.922 using the full training dataset. The smaller model, Flan-T5-xl, requires fine-tuning with only 2.3M additional parameters to achieve comparable performance to the fully fine-tuned Gatortron-base model, both surpassing 0.9 F1 score. Conclusion: Open-source instruction-tuned LLMs demonstrate impressive in-context learning capability in the mobility functioning classification task. The performance of these models can be further improved by continuing fine-tuning on a task-specific dataset.