Multi-modal contrastive learning as a self-supervised representation learning technique has achieved great success in foundation model training, such as CLIP~\citep{radford2021learning}. In this paper, we study the theoretical properties of the learned representations from multi-modal contrastive learning beyond linear representations and specific data distributions. Our analysis reveals that, enabled by temperature optimization, multi-modal contrastive learning not only maximizes mutual information between modalities but also adapts to intrinsic dimensions of data, which can be much lower than user-specified dimensions for representation vectors. Experiments on both synthetic and real-world datasets demonstrate the ability of contrastive learning to learn low-dimensional and informative representations, bridging theoretical insights and practical performance.

Before deploying outputs from foundation models in high-stakes tasks, it is imperative to ensure that they align with human values. For instance, in radiology report generation, reports generated by a vision-language model must align with human evaluations before their use in medical decision-making. This paper presents Conformal Alignment, a general framework for identifying units whose outputs meet a user-specified alignment criterion. It is guaranteed that on average, a prescribed fraction of selected units indeed meet the alignment criterion, regardless of the foundation model or the data distribution. Given any pre-trained model and new units with model-generated outputs, Conformal Alignment leverages a set of reference data with ground-truth alignment status to train an alignment predictor. It then selects new units whose predicted alignment scores surpass a data-dependent threshold, certifying their corresponding outputs as trustworthy. Through applications to question answering and radiology report generation, we demonstrate that our method is able to accurately identify units with trustworthy outputs via lightweight training over a moderate amount of reference data. En route, we investigate the informativeness of various features in alignment prediction and combine them with standard models to construct the alignment predictor.

Matrix completion aims to estimate missing entries in a data matrix, using the assumption of a low-complexity structure (e.g., low-rankness) so that imputation is possible. While many effective estimation algorithms exist in the literature, uncertainty quantification for this problem has proved to be challenging, and existing methods are extremely sensitive to model misspecification. In this work, we propose a distribution-free method for predictive inference in the matrix completion problem. Our method adapts the framework of conformal prediction, which provides prediction intervals with guaranteed distribution-free validity in the setting of regression, to the problem of matrix completion. Our resulting method, conformalized matrix completion (cmc), offers provable predictive coverage regardless of the accuracy of the low-rank model. Empirical results on simulated and real data demonstrate that cmc is robust to model misspecification while matching the performance of existing model-based methods when the model is correct.

Spiking neural network (SNN) can process discrete spikes and offers a high degree of real-time performance and excellent energy efficiency ratio. However, most current neuromorphic hardware platforms lack efficient driven algorithms and only support a single type of neuron model, which has slow speed and poor scalability. This paper proposes a general FPGA-based neuromorphic hardware acceleration platform (FPGA-NHAP), supporting the effective inference and acceleration of SNN network with low power, high speed and good scalability. First, a neuron computing unit is designed to simulate the both LIF and Izhikevich (IZH) neurons with the parallel spike caching and scheduling technique. Second, a novel integrated driven update algorithm is proposed to complete the spike encoding of external data, reducing the waiting time of neuron state update effectively. Third, the proposed platform is implemented using a RISC-V processor and a Xilinx FPGA, simulating 16,384 neurons and 16.8 million synapses with a power consumption of 0.535 W. Finally, two different three-layer SNN networks are deployed on the proposed platform for recognition tasks on the MNIST and Fashion-MNIST datasets, achieving the accuracy of 97.70%, 85.14% (LIF) and 97.81%, 83.16% (IZH), frame rates of 208 frame/s, 128 frame/s (LIF) and 206 frame/s, 141 frame/s (IZH), respectively.

Deep neural networks (DNNs) have become increasingly popular and achieved outstanding performance in predictive tasks. However, the DNN framework itself cannot inform the user which features are more or less relevant for making the prediction, which limits its applicability in many scientific fields. We introduce neural Gaussian mirrors (NGMs), in which mirrored features are created, via a structured perturbation based on a kernel-based conditional dependence measure, to help evaluate feature importance. We design two modifications of the DNN architecture for incorporating mirrored features and providing mirror statistics to measure feature importance. As shown in simulated and real data examples, the proposed method controls the feature selection error rate at a predefined level and maintains a high selection power even with the presence of highly correlated features.