We summarized both common and novel predictive models used for stock price prediction and combined them with technical indices, fundamental characteristics and text-based sentiment data to predict S&P stock prices. A 66.18% accuracy in S&P 500 index directional prediction and 62.09% accuracy in individual stock directional prediction was achieved by combining different machine learning models such as Random Forest and LSTM together into state-of-the-art ensemble models. The data we use contains weekly historical prices, finance reports, and text information from news items associated with 518 different common stocks issued by current and former S&P 500 large-cap companies, from January 1, 2000 to December 31, 2019. Our study's innovation includes utilizing deep language models to categorize and infer financial news item sentiment; fusing different models containing different combinations of variables and stocks to jointly make predictions; and overcoming the insufficient data problem for machine learning models in time series by using data across different stocks.

The physical world is governed by the laws of physics, often represented in form of nonlinear partial differential equations (PDEs). Unfortunately, solution of PDEs is non-trivial and often involves significant computational time. With recent developments in the field of artificial intelligence and machine learning, the solution of PDEs using neural network has emerged as a domain with huge potential. However, most of the developments in this field are based on either fully connected neural networks (FNN) or convolutional neural networks (CNN). While FNN is computationally inefficient as the number of network parameters can be potentially huge, CNN necessitates regular grid and simpler domain. In this work, we propose a novel framework referred to as the Graph Attention Differential Equation (GrADE) for solving time dependent nonlinear PDEs. The proposed approach couples FNN, graph neural network, and recently developed Neural ODE framework. The primary idea is to use graph neural network for modeling the spatial domain, and Neural ODE for modeling the temporal domain. The attention mechanism identifies important inputs/features and assign more weightage to the same; this enhances the performance of the proposed framework. Neural ODE, on the other hand, results in constant memory cost and allows trading of numerical precision for speed. We also propose depth refinement as an effective technique for training the proposed architecture in lesser time with better accuracy. The effectiveness of the proposed framework is illustrated using 1D and 2D Burgers' equations. Results obtained illustrate the capability of the proposed framework in modeling PDE and its scalability to larger domains without the need for retraining.

This paper identifies a structural property of data distributions that enables deep neural networks to learn hierarchically. We define the "staircase" property for functions over the Boolean hypercube, which posits that high-order Fourier coefficients are reachable from lower-order Fourier coefficients along increasing chains. We prove that functions satisfying this property can be learned in polynomial time using layerwise stochastic coordinate descent on regular neural networks -- a class of network architectures and initializations that have homogeneity properties. Our analysis shows that for such staircase functions and neural networks, the gradient-based algorithm learns high-level features by greedily combining lower-level features along the depth of the network. We further back our theoretical results with experiments showing that staircase functions are also learnable by more standard ResNet architectures with stochastic gradient descent. Both the theoretical and experimental results support the fact that staircase properties have a role to play in understanding the capabilities of gradient-based learning on regular networks, in contrast to general polynomial-size networks that can emulate any SQ or PAC algorithms as recently shown.

Multiclass multilabel classification refers to the task of attributing multiple labels to examples via predictions. Current models formulate a reduction of that multilabel setting into either multiple binary classifications or multiclass classification, allowing for the use of existing loss functions (sigmoid, cross-entropy, logistic, etc.). Empirically, these methods have been reported to achieve good performance on different metrics (F1 score, Recall, Precision, etc.). Theoretically though, the multilabel classification reductions does not accommodate for the prediction of varying numbers of labels per example and the underlying losses are distant estimates of the performance metrics. We propose a loss function, sigmoidF1. It is an approximation of the F1 score that (I) is smooth and tractable for stochastic gradient descent, (II) naturally approximates a multilabel metric, (III) estimates label propensities and label counts. More generally, we show that any confusion matrix metric can be formulated with a smooth surrogate. We evaluate the proposed loss function on different text and image datasets, and with a variety of metrics, to account for the complexity of multilabel classification evaluation. In our experiments, we embed the sigmoidF1 loss in a classification head that is attached to state-of-the-art efficient pretrained neural networks MobileNetV2 and DistilBERT. Our experiments show that sigmoidF1 outperforms other loss functions on four datasets and several metrics. These results show the effectiveness of using inference-time metrics as loss function at training time in general and their potential on non-trivial classification problems like multilabel classification.

It is appealing but challenging to achieve real-time deep neural network (DNN) inference on mobile devices because even the powerful modern mobile devices are considered as ``resource-constrained'' when executing large-scale DNNs. It necessitates the sparse model inference via weight pruning, i.e., DNN weight sparsity, and it is desirable to design a new DNN weight sparsity scheme that can facilitate real-time inference on mobile devices while preserving a high sparse model accuracy. This paper designs a novel mobile inference acceleration framework GRIM that is General to both convolutional neural networks (CNNs) and recurrent neural networks (RNNs) and that achieves Real-time execution and high accuracy, leveraging fine-grained structured sparse model Inference and compiler optimizations for Mobiles. We start by proposing a new fine-grained structured sparsity scheme through the Block-based Column-Row (BCR) pruning. Based on this new fine-grained structured sparsity, our GRIM framework consists of two parts: (a) the compiler optimization and code generation for real-time mobile inference; and (b) the BCR pruning optimizations for determining pruning hyperparameters and performing weight pruning. We compare GRIM with Alibaba MNN, TVM, TensorFlow-Lite, a sparse implementation based on CSR, PatDNN, and ESE (a representative FPGA inference acceleration framework for RNNs), and achieve up to 14.08x speedup.

Breast histology image classification is a crucial step in the early diagnosis of breast cancer. In breast pathological diagnosis, Convolutional Neural Networks (CNNs) have demonstrated great success using digitized histology slides. However, tissue classification is still challenging due to the high visual variability of the large-sized digitized samples and the lack of contextual information. In this paper, we propose a novel CNN, called Multi-level Context and Uncertainty aware (MCUa) dynamic deep learning ensemble model.MCUamodel consists of several multi-level context-aware models to learn the spatial dependency between image patches in a layer-wise fashion. It exploits the high sensitivity to the multi-level contextual information using an uncertainty quantification component to accomplish a novel dynamic ensemble model.MCUamodelhas achieved a high accuracy of 98.11% on a breast cancer histology image dataset. Experimental results show the superior effectiveness of the proposed solution compared to the state-of-the-art histology classification models.

In recent years, deep learning-based methods have shown promising results in computer vision area. However, a common deep learning model requires a large amount of labeled data, which is labor-intensive to collect and label. What's more, the model can be ruined due to the domain shift between training data and testing data. Text recognition is a broadly studied field in computer vision and suffers from the same problems noted above due to the diversity of fonts and complicated backgrounds. In this paper, we focus on the text recognition problem and mainly make three contributions toward these problems. First, we collect a multi-source domain adaptation dataset for text recognition, including five different domains with over five million images, which is the first multi-domain text recognition dataset to our best knowledge. Secondly, we propose a new method called Meta Self-Learning, which combines the self-learning method with the meta-learning paradigm and achieves a better recognition result under the scene of multi-domain adaptation. Thirdly, extensive experiments are conducted on the dataset to provide a benchmark and also show the effectiveness of our method. The code of our work and dataset are available soon at https://bupt-ai-cz.github.io/Meta-SelfLearning/.

In this thesis, we introduce Greenformers, a collection of model efficiency methods to improve the model efficiency of the recently renowned transformer models with a low-rank approximation approach. The development trend of deep learning models tends to results in a more complex and larger model. Although it leads to a better and more accurate prediction, the resulting model becomes even more costly, as it requires weeks of training with a huge amount of GPU resources. Particularly, the size and computational cost of transformer-based models have increased tremendously since its first debut in 2017 from ~100 million parameters up to ~1.6 trillion parameters in early 2021. This computationally hungry model also incurs a substantial cost to the environment and even reaches an alarming level of carbon footprint. Some of these models are so massive that it is even impossible to run the model without a GPU cluster. Greenformers improve the model efficiency of transformer models by applying low-rank approximation approaches. Specifically, we propose a low-rank factorization approach to improve the efficiency of the transformer model called Low-Rank Transformer. We further compare our model with an existing low-rank factorization approach called Linformer. Based on our analysis, the Low-Rank Transformer model is suitable for improving both the time and memory efficiency in processing short-sequence (<= 512) input data, while the Linformer model is suitable for improving the efficiency in processing long-sequence input data (>= 512). We also show that Low-Rank Transformer is more suitable for on-device deployment, as it significantly reduces the model size. Additionally, we estimate that applying LRT to the existing BERT-base model can significantly reduce the computational, economical, and environmental costs for developing such models by more than 30% of its original costs.

When recurrent neural network transducers (RNNTs) are trained using the typical maximum likelihood criterion, the prediction network is trained only on ground truth label sequences. This leads to a mismatch during inference, known as exposure bias, when the model must deal with label sequences containing errors. In this paper we investigate approaches to reducing exposure bias in training to improve the generalization of RNNT models for automatic speech recognition (ASR). A label-preserving input perturbation to the prediction network is introduced. The input token sequences are perturbed using SwitchOut and scheduled sampling based on an additional token language model. Experiments conducted on the 300-hour Switchboard dataset demonstrate their effectiveness. By reducing the exposure bias, we show that we can further improve the accuracy of a high-performance RNNT ASR model and obtain state-of-the-art results on the 300-hour Switchboard dataset.

Privacy protection is an ethical issue with broad concern in Artificial Intelligence (AI). Federated learning is a new machine learning paradigm to learn a shared model across users or organisations without direct access to the data. It has great potential to be the next-general AI model training framework that offers privacy protection and therefore has broad implications for the future of digital health and healthcare informatics. Implementing an open innovation framework in the healthcare industry, namely open health, is to enhance innovation and creative capability of health-related organisations by building a next-generation collaborative framework with partner organisations and the research community. In particular, this game-changing collaborative framework offers knowledge sharing from diverse data with a privacy-preserving. This chapter will discuss how federated learning can enable the development of an open health ecosystem with the support of AI. Existing challenges and solutions for federated learning will be discussed.