Understanding Neural Network Binarization with Forward and Backward Proximal Quantizers

In neural network binarization, BinaryConnect (BC) and its variants are considered the standard. These methods apply the sign function in their forward pass and their respective gradients are backpropagated to update the weights. However, the derivative of the sign function is zero whenever defined, which consequently freezes training. Therefore, implementations of BC (e.g., BNN) usually replace the derivative of sign in the backward computation with identity or other approximate gradient alternatives. Although such practice works well empirically, it is largely a heuristic or ''training trick.'' We aim at shedding some light on these training tricks from the optimization perspective. Building from existing theory on ProxConnect (PC, a generalization of BC), we (1) equip PC with different forward-backward quantizers and obtain ProxConnect++ (PC++) that includes existing binarization techniques as special cases; (2) derive a principled way to synthesize forward-backward quantizers with automatic theoretical guarantees; (3) illustrate our theory by proposing an enhanced binarization algorithm BNN++; (4) conduct image classification experiments on CNNs and vision transformers, and empirically verify that BNN++ generally achieves competitive results on binarizing these models.

BinaryViT: Pushing Binary Vision Transformers Towards Convolutional Models

With the increasing popularity and the increasing size of vision transformers (ViTs), there has been an increasing interest in making them more efficient and less computationally costly for deployment on edge devices with limited computing resources. Binarization can be used to help reduce the size of ViT models and their computational cost significantly, using popcount operations when the weights and the activations are in binary. However, ViTs suffer a larger performance drop when directly applying convolutional neural network (CNN) binarization methods or existing binarization methods to binarize ViTs compared to CNNs on datasets with a large number of classes such as ImageNet-1k. With extensive analysis, we find that binary vanilla ViTs such as DeiT miss out on a lot of key architectural properties that CNNs have that allow binary CNNs to have much higher representational capability than binary vanilla ViT. Therefore, we propose BinaryViT, in which inspired by the CNN architecture, we include operations from the CNN architecture into a pure ViT architecture to enrich the representational capability of a binary ViT without introducing convolutions. These include an average pooling layer instead of a token pooling layer, a block that contains multiple average pooling branches, an affine transformation right before the addition of each main residual connection, and a pyramid structure. Experimental results on the ImageNet-1k dataset show the effectiveness of these operations that allow a binary pure ViT model to be competitive with previous state-of-the-art (SOTA) binary CNN models.

Mathematical Challenges in Deep Learning

Deep models are dominating the artificial intelligence (AI) industry since the ImageNet challenge in 2012. The size of deep models is increasing ever since, which brings new challenges to this field with applications in cell phones, personal computers, autonomous cars, and wireless base stations. Here we list a set of problems, ranging from training, inference, generalization bound, and optimization with some formalism to communicate these challenges with mathematicians, statisticians, and theoretical computer scientists. This is a subjective view of the research questions in deep learning that benefits the tech industry in long run.

EuclidNets: An Alternative Operation for Efficient Inference of Deep Learning Models

With the advent of deep learning application on edge devices, researchers actively try to optimize their deployments on low-power and restricted memory devices. There are established compression method such as quantization, pruning, and architecture search that leverage commodity hardware. Apart from conventional compression algorithms, one may redesign the operations of deep learning models that lead to more efficient implementation. To this end, we propose EuclidNet, a compression method, designed to be implemented on hardware which replaces multiplication, $xw$, with Euclidean distance $(x-w)^2$. We show that EuclidNet is aligned with matrix multiplication and it can be used as a measure of similarity in case of convolutional layers. Furthermore, we show that under various transformations and noise scenarios, EuclidNet exhibits the same performance compared to the deep learning models designed with multiplication operations.

DenseShift: Towards Accurate and Transferable Low-Bit Shift Network

Deploying deep neural networks on low-resource edge devices is challenging due to their ever-increasing resource requirements. Recent investigations propose multiplication-free neural networks to reduce computation and memory consumption. Shift neural network is one of the most effective tools towards these reductions. However, existing low-bit shift networks are not as accurate as their full precision counterparts and cannot efficiently transfer to a wide range of tasks due to their inherent design flaws. We propose DenseShift network that exploits the following novel designs. First, we demonstrate that the zero-weight values in low-bit shift networks are neither useful to the model capacity nor simplify the model inference. Therefore, we propose to use a zero-free shifting mechanism to simplify inference while increasing the model capacity. Second, we design a new metric to measure the weight freezing issue in training low-bit shift networks, and propose a sign-scale decomposition to improve the training efficiency. Third, we propose the low-variance random initialization strategy to improve the model's performance in transfer learning scenarios. We run extensive experiments on various computer vision and speech tasks. The experimental results show that DenseShift network significantly outperforms existing low-bit multiplication-free networks and can achieve competitive performance to the full-precision counterpart. It also exhibits strong transfer learning performance with no drop in accuracy.

Low-bit Shift Network for End-to-End Spoken Language Understanding

Deep neural networks (DNN) have achieved impressive success in multiple domains. Over the years, the accuracy of these models has increased with the proliferation of deeper and more complex architectures. Thus, state-of-the-art solutions are often computationally expensive, which makes them unfit to be deployed on edge computing platforms. In order to mitigate the high computation, memory, and power requirements of inferring convolutional neural networks (CNNs), we propose the use of power-of-two quantization, which quantizes continuous parameters into low-bit power-of-two values. This reduces computational complexity by removing expensive multiplication operations and with the use of low-bit weights. ResNet is adopted as the building block of our solution and the proposed model is evaluated on a spoken language understanding (SLU) task. Experimental results show improved performance for shift neural network architectures, with our low-bit quantization achieving 98.76 \% on the test set which is comparable performance to its full-precision counterpart and state-of-the-art solutions.

Deep Neural Networks pruning via the Structured Perspective Regularization

In Machine Learning, Artificial Neural Networks (ANNs) are a very powerful tool, broadly used in many applications. Often, the selected (deep) architectures include many layers, and therefore a large amount of parameters, which makes training, storage and inference expensive. This motivated a stream of research about compressing the original networks into smaller ones without excessively sacrificing performances. Among the many proposed compression approaches, one of the most popular is \emph{pruning}, whereby entire elements of the ANN (links, nodes, channels, \ldots) and the corresponding weights are deleted. Since the nature of the problem is inherently combinatorial (what elements to prune and what not), we propose a new pruning method based on Operational Research tools. We start from a natural Mixed-Integer-Programming model for the problem, and we use the Perspective Reformulation technique to strengthen its continuous relaxation. Projecting away the indicator variables from this reformulation yields a new regularization term, which we call the Structured Perspective Regularization, that leads to structured pruning of the initial architecture. We test our method on some ResNet architectures applied to CIFAR-10, CIFAR-100 and ImageNet datasets, obtaining competitive performances w.r.t.~the state of the art for structured pruning.

$S^3$: Sign-Sparse-Shift Reparametrization for Effective Training of Low-bit Shift Networks

Shift neural networks reduce computation complexity by removing expensive multiplication operations and quantizing continuous weights into low-bit discrete values, which are fast and energy efficient compared to conventional neural networks. However, existing shift networks are sensitive to the weight initialization, and also yield a degraded performance caused by vanishing gradient and weight sign freezing problem. To address these issues, we propose S low-bit re-parameterization, a novel technique for training low-bit shift networks. Our method decomposes a discrete parameter in a sign-sparse-shift 3-fold manner. In this way, it efficiently learns a low-bit network with a weight dynamics similar to full-precision networks and insensitive to weight initialization. Our proposed training method pushes the boundaries of shift neural networks and shows 3-bit shift networks out-performs their full-precision counterparts in terms of top-1 accuracy on ImageNet.

Tensor train decompositions on recurrent networks

Recurrent neural networks (RNN) such as long-short-term memory (LSTM) networks are essential in a multitude of daily live tasks such as speech, language, video, and multimodal learning. The shift from cloud to edge computation intensifies the need to contain the growth of RNN parameters. Current research on RNN shows that despite the performance obtained on convolutional neural networks (CNN), keeping a good performance in compressed RNNs is still a challenge. Most of the literature on compression focuses on CNNs using matrix product (MPO) operator tensor trains. However, matrix product state (MPS) tensor trains have more attractive features than MPOs, in terms of storage reduction and computing time at inference. We show that MPS tensor trains should be at the forefront of LSTM network compression through a theoretical analysis and practical experiments on NLP task.

Clustering Causal Additive Noise Models

Additive noise models are commonly used to infer the causal direction for a given set of observed data. Most causal models assume a single homogeneous population. However, observations may be collected under different conditions in practice. Such data often require models that can accommodate possible heterogeneity caused by different conditions under which data have been collected. We propose a clustering algorithm inspired by the $k$-means algorithm, but with unknown $k$. Using the proposed algorithm, both the labels and the number of components are estimated from the collected data. The estimated labels are used to adjust the causal direction test statistic. The adjustment significantly improves the performance of the test statistic in identifying the correct causal direction.