Mitigating the climate crisis requires a rapid transition towards lower-carbon energy. Catalyst materials play a crucial role in the electrochemical reactions involved in numerous industrial processes key to this transition, such as renewable energy storage and electrofuel synthesis. To reduce the energy spent on such activities, we must quickly discover more efficient catalysts to drive electrochemical reactions. Machine learning (ML) holds the potential to efficiently model materials properties from large amounts of data, accelerating electrocatalyst design. The Open Catalyst Project OC20 dataset was constructed to that end. However, ML models trained on OC20 are still neither scalable nor accurate enough for practical applications. In this paper, we propose task-specific innovations applicable to most architectures, enhancing both computational efficiency and accuracy. This includes improvements in (1) the graph creation step, (2) atom representations, (3) the energy prediction head, and (4) the force prediction head. We describe these contributions, referred to as PhAST, and evaluate them thoroughly on multiple architectures. Overall, PhAST improves energy MAE by 4 to 42$\%$ while dividing compute time by 3 to 8$\times$ depending on the targeted task/model. PhAST also enables CPU training, leading to 40$\times$ speedups in highly parallelized settings. Python package: \url{https://phast.readthedocs.io}.
Decades of research on the 0-1 knapsack problem led to very efficient algorithms that are able to quickly solve large problem instances to optimality. This prompted researchers to also investigate whether relatively small problem instances exist that are hard for existing solvers and investigate which features characterize their hardness. Previously the authors proposed a new class of hard 0-1 knapsack problem instances and demonstrated that the properties of so-called inclusionwise maximal solutions (IMSs) can be important hardness indicators for this class. In the current paper, we formulate several new computationally challenging problems related to the IMSs of arbitrary 0-1 knapsack problem instances. Based on generalizations of previous work and new structural results about IMSs, we formulate polynomial and pseudopolynomial time algorithms for solving these problems. From this we derive a set of 14 computationally expensive features, which we calculate for two large datasets on a supercomputer in approximately 540 CPU-hours. We show that the proposed features contain important information related to the empirical hardness of a problem instance that was missing in earlier features from the literature by training machine learning models that can accurately predict the empirical hardness of a wide variety of 0-1 knapsack problem instances. Using the instance space analysis methodology, we also show that hard 0-1 knapsack problem instances are clustered together around a relatively dense region of the instance space and several features behave differently in the easy and hard parts of the instance space.
From AlexNet to Inception, autoencoders to diffusion models, the development of novel and powerful deep learning models and learning algorithms has proceeded at breakneck speeds. In part, we believe that rapid iteration of model architecture and learning techniques by a large community of researchers over a common representation of the underlying entities has resulted in transferable deep learning knowledge. As a result, model scale, accuracy, fidelity, and compute performance have dramatically increased in computer vision and natural language processing. On the other hand, the lack of a common representation for chemical structure has hampered similar progress. To enable transferable deep learning, we identify the need for a robust 3-dimensional representation of materials such as molecules and crystals. The goal is to enable both materials property prediction and materials generation with 3D structures. While computationally costly, such representations can model a large set of chemical structures. We propose $\textit{ParticleGrid}$, a SIMD-optimized library for 3D structures, that is designed for deep learning applications and to seamlessly integrate with deep learning frameworks. Our highly optimized grid generation allows for generating grids on the fly on the CPU, reducing storage and GPU compute and memory requirements. We show the efficacy of 3D grids generated via $\textit{ParticleGrid}$ and accurately predict molecular energy properties using a 3D convolutional neural network. Our model is able to get 0.006 mean square error and nearly match the values calculated using computationally costly density functional theory at a fraction of the time.
Modern IT system operation demands the integration of system software and hardware metrics. As a result, it generates a massive amount of data, which can be potentially used to make data-driven operational decisions. In the basic form, the decision model needs to monitor a large set of machine data, such as CPU utilization, allocated memory, disk and network latency, and predicts the system metrics to prevent performance degradation. Nevertheless, building an effective prediction model in this scenario is rather challenging as the model has to accurately capture the long-range coupling dependency in the Multivariate Time-Series (MTS). Moreover, this model needs to have low computational complexity and can scale efficiently to the dimension of data available. In this paper, we propose a highly efficient model named HigeNet to predict the long-time sequence time series. We have deployed the HigeNet on production in the D-matrix platform. We also provide offline evaluations on several publicly available datasets as well as one online dataset to demonstrate the model's efficacy. The extensive experiments show that training time, resource usage and accuracy of the model are found to be significantly better than five state-of-the-art competing models.
Unpacking and comprehending how black-box machine learning algorithms make decisions has been a persistent challenge for researchers and end-users. Explaining time-series predictive models is useful for clinical applications with high stakes to understand the behavior of prediction models. However, existing approaches to explain such models are frequently unique to data where the features do not have a time-varying component. In this paper, we introduce WindowSHAP, a model-agnostic framework for explaining time-series classifiers using Shapley values. We intend for WindowSHAP to mitigate the computational complexity of calculating Shapley values for long time-series data as well as improve the quality of explanations. WindowSHAP is based on partitioning a sequence into time windows. Under this framework, we present three distinct algorithms of Stationary, Sliding and Dynamic WindowSHAP, each evaluated against baseline approaches, KernelSHAP and TimeSHAP, using perturbation and sequence analyses metrics. We applied our framework to clinical time-series data from both a specialized clinical domain (Traumatic Brain Injury - TBI) as well as a broad clinical domain (critical care medicine). The experimental results demonstrate that, based on the two quantitative metrics, our framework is superior at explaining clinical time-series classifiers, while also reducing the complexity of computations. We show that for time-series data with 120 time steps (hours), merging 10 adjacent time points can reduce the CPU time of WindowSHAP by 80% compared to KernelSHAP. We also show that our Dynamic WindowSHAP algorithm focuses more on the most important time steps and provides more understandable explanations. As a result, WindowSHAP not only accelerates the calculation of Shapley values for time-series data, but also delivers more understandable explanations with higher quality.
Recent years have witnessed the booming of various differentiable optimization algorithms. These algorithms exhibit different execution patterns, and their execution needs massive computational resources that go beyond a single CPU and GPU. Existing differentiable optimization libraries, however, cannot support efficient algorithm development and multi-CPU/GPU execution, making the development of differentiable optimization algorithms often cumbersome and expensive. This paper introduces TorchOpt, a PyTorch-based efficient library for differentiable optimization. TorchOpt provides a unified and expressive differentiable optimization programming abstraction. This abstraction allows users to efficiently declare and analyze various differentiable optimization programs with explicit gradients, implicit gradients, and zero-order gradients. TorchOpt further provides a high-performance distributed execution runtime. This runtime can fully parallelize computation-intensive differentiation operations (e.g. tensor tree flattening) on CPUs / GPUs and automatically distribute computation to distributed devices. Experimental results show that TorchOpt achieves $5.2\times$ training time speedup on an 8-GPU server. TorchOpt is available at: https://github.com/metaopt/torchopt/.
Protein structure prediction is a critical problem linked to drug design, mutation detection, and protein synthesis, among other applications. To this end, evolutionary data has been used to build contact maps which are traditionally minimized as energy functions via gradient descent based schemes like the L-BFGS algorithm. In this paper we present what we call the Alternating Metropolis-Hastings (AMH) algorithm, which (a) significantly improves the performance of traditional MCMC methods, (b) is inherently parallelizable allowing significant hardware acceleration using GPU, and (c) can be integrated with the L-BFGS algorithm to improve its performance. The algorithm shows an improvement in energy of found structures of 8.17% to 61.04% (average 38.9%) over traditional MH and 0.53% to 17.75% (average 8.9%) over traditional MH with intermittent noisy restarts, tested across 9 proteins from recent CASP competitions. We go on to map the Alternating MH algorithm to a GPGPU which improves sampling rate by 277x and improves simulation time to a low energy protein prediction by 7.5x to 26.5x over CPU. We show that our approach can be incorporated into state-of-the-art protein prediction pipelines by applying it to both trRosetta2's energy function and the distogram component of Alphafold1's energy function. Finally, we note that specially designed probabilistic computers (or p-computers) can provide even better performance than GPUs for MCMC algorithms like the one discussed here.
Tensor program tuning is a non-convex objective optimization problem, to which search-based approaches have proven to be effective. At the core of the search-based approaches lies the design of the cost model. Though deep learning-based cost models perform significantly better than other methods, they still fall short and suffer from the following problems. First, their feature extraction heavily relies on expert-level domain knowledge in hardware architectures. Even so, the extracted features are often unsatisfactory and require separate considerations for CPUs and GPUs. Second, a cost model trained on one hardware platform usually performs poorly on another, a problem we call cross-hardware unavailability. In order to address these problems, we propose TLP and MTLTLP. TLP is a deep learning-based cost model that facilitates tensor program tuning. Instead of extracting features from the tensor program itself, TLP extracts features from the schedule primitives. We treat schedule primitives as tensor languages. TLP is thus a Tensor Language Processing task. In this way, the task of predicting the tensor program latency through the cost model is transformed into a natural language processing (NLP) regression task. MTL-TLP combines Multi-Task Learning and TLP to cope with the cross-hardware unavailability problem. We incorporate these techniques into the Ansor framework and conduct detailed experiments. Results show that TLP can speed up the average search time by 9.1X and 3.0X on CPU and GPU workloads, respectively, compared to the state-of-the-art implementation. MTL-TLP can achieve a speed-up of 4.7X and 2.9X on CPU and GPU workloads, respectively, using only 7% of the target hardware data.
Mutation-based fuzzing has become one of the most common vulnerability discovery solutions over the last decade. Fuzzing can be optimized when targeting specific programs, and given that, some studies have employed online optimization methods to do it automatically, i.e., tuning fuzzers for any given program in a program-agnostic manner. However, previous studies have neither fully explored mutation schemes suitable for online optimization methods, nor online optimization methods suitable for mutation schemes. In this study, we propose an optimization framework called SLOPT that encompasses both a bandit-friendly mutation scheme and mutation-scheme-friendly bandit algorithms. The advantage of SLOPT is that it can generally be incorporated into existing fuzzers, such as AFL and Honggfuzz. As a proof of concept, we implemented SLOPT-AFL++ by integrating SLOPT into AFL++ and showed that the program-agnostic optimization delivered by SLOPT enabled SLOPT-AFL++ to achieve higher code coverage than AFL++ in all of ten real-world FuzzBench programs. Moreover, we ran SLOPT-AFL++ against several real-world programs from OSS-Fuzz and successfully identified three previously unknown vulnerabilities, even though these programs have been fuzzed by AFL++ for a considerable number of CPU days on OSS-Fuzz.
CatBoost is a popular machine learning library. CatBoost models are based on oblivious decision trees, making training and evaluation rapid. CatBoost has many applications, and some require low latency and high throughput evaluation. This paper investigates the possibilities for improving CatBoost's performance in single-core CPU computations. We explore the new features provided by the AVX instruction sets to optimize evaluation. We increase performance by 20-40% using AVX2 instructions without quality impact. We also introduce a new trade-off between speed and quality. Using float16 for leaf values and AVX-512 instructions, we achieve 50-70% speed-up.
Due to their high temporal resolution and large dynamic range event cameras are uniquely suited for the analysis of time-periodic signals in an image. In this work we present an efficient and fully asynchronous event camera algorithm for detecting the fundamental frequency at which image pixels flicker. The algorithm employs a second-order digital infinite impulse response (IIR) filter to perform an approximate per-pixel brightness reconstruction and is more robust to high-frequency noise than the baseline method we compare to. We further demonstrate that using the falling edge of the signal leads to more accurate period estimates than the rising edge, and that for certain signals interpolating the zero-level crossings can further increase accuracy. Our experiments find that the outstanding capabilities of the camera in detecting frequencies up to 64kHz for a single pixel do not carry over to full sensor imaging as readout bandwidth limitations become a serious obstacle. This suggests that a hardware implementation closer to the sensor will allow for greatly improved frequency imaging. We discuss the important design parameters for fullsensor frequency imaging and present Frequency Cam, an open-source implementation as a ROS node that can run on a single core of a laptop CPU at more than 50 million events per second. It produces results that are qualitatively very similar to those obtained from the closed source vibration analysis module in Prophesee's Metavision Toolkit. The code for Frequency Cam and a demonstration video can be found at https://github.com/berndpfrommer/frequency_cam
We present the Open MatSci ML Toolkit: a flexible, self-contained, and scalable Python-based framework to apply deep learning models and methods on scientific data with a specific focus on materials science and the OpenCatalyst Dataset. Our toolkit provides: 1. A scalable machine learning workflow for materials science leveraging PyTorch Lightning, which enables seamless scaling across different computation capabilities (laptop, server, cluster) and hardware platforms (CPU, GPU, XPU). 2. Deep Graph Library (DGL) support for rapid graph neural network prototyping and development. By publishing and sharing this toolkit with the research community via open-source release, we hope to: 1. Lower the entry barrier for new machine learning researchers and practitioners that want to get started with the OpenCatalyst dataset, which presently comprises the largest computational materials science dataset. 2. Enable the scientific community to apply advanced machine learning tools to high-impact scientific challenges, such as modeling of materials behavior for clean energy applications. We demonstrate the capabilities of our framework by enabling three new equivariant neural network models for multiple OpenCatalyst tasks and arrive at promising results for compute scaling and model performance.
We propose a lightweight end-to-end text-to-speech model using multi-band generation and inverse short-time Fourier transform. Our model is based on VITS, a high-quality end-to-end text-to-speech model, but adopts two changes for more efficient inference: 1) the most computationally expensive component is partially replaced with a simple inverse short-time Fourier transform, and 2) multi-band generation, with fixed or trainable synthesis filters, is used to generate waveforms. Unlike conventional lightweight models, which employ optimization or knowledge distillation separately to train two cascaded components, our method enjoys the full benefits of end-to-end optimization. Experimental results show that our model synthesized speech as natural as that synthesized by VITS, while achieving a real-time factor of 0.066 on an Intel Core i7 CPU, 4.1 times faster than VITS. Moreover, a smaller version of the model significantly outperformed a lightweight baseline model with respect to both naturalness and inference speed. Code and audio samples are available from https://github.com/MasayaKawamura/MB-iSTFT-VITS.
While recent research advances in speaker diarization mostly focus on improving the quality of diarization results, there is also an increasing interest in improving the efficiency of diarization systems. In this paper, we demonstrate that a multi-stage clustering strategy that uses different clustering algorithms for input of different lengths can address multi-faceted challenges of on-device speaker diarization applications. Specifically, a fallback clusterer is used to handle short-form inputs; a main clusterer is used to handle medium-length inputs; and a pre-clusterer is used to compress long-form inputs before they are processed by the main clusterer. Both the main clusterer and the pre-clusterer can be configured with an upper bound of the computational complexity to adapt to devices with different resource constraints. This multi-stage clustering strategy is critical for streaming on-device speaker diarization systems, where the budgets of CPU, memory and battery are tight.
Recently, research has increasingly focused on developing efficient neural network architectures. In this work, we explore logic gate networks for machine learning tasks by learning combinations of logic gates. These networks comprise logic gates such as "AND" and "XOR", which allow for very fast execution. The difficulty in learning logic gate networks is that they are conventionally non-differentiable and therefore do not allow training with gradient descent. Thus, to allow for effective training, we propose differentiable logic gate networks, an architecture that combines real-valued logics and a continuously parameterized relaxation of the network. The resulting discretized logic gate networks achieve fast inference speeds, e.g., beyond a million images of MNIST per second on a single CPU core.
Recommendation systems are of crucial importance for a variety of modern apps and web services, such as news feeds, social networks, e-commerce, search, etc. To achieve peak prediction accuracy, modern recommendation models combine deep learning with terabyte-scale embedding tables to obtain a fine-grained representation of the underlying data. Traditional inference serving architectures require deploying the whole model to standalone servers, which is infeasible at such massive scale. In this paper, we provide insights into the intriguing and challenging inference domain of online recommendation systems. We propose the HugeCTR Hierarchical Parameter Server (HPS), an industry-leading distributed recommendation inference framework, that combines a high-performance GPU embedding cache with an hierarchical storage architecture, to realize low-latency retrieval of embeddings for online model inference tasks. Among other things, HPS features (1) a redundant hierarchical storage system, (2) a novel high-bandwidth cache to accelerate parallel embedding lookup on NVIDIA GPUs, (3) online training support and (4) light-weight APIs for easy integration into existing large-scale recommendation workflows. To demonstrate its capabilities, we conduct extensive studies using both synthetically engineered and public datasets. We show that our HPS can dramatically reduce end-to-end inference latency, achieving 5~62x speedup (depending on the batch size) over CPU baseline implementations for popular recommendation models. Through multi-GPU concurrent deployment, the HPS can also greatly increase the inference QPS.
Designing feasible and effective architectures under diverse computational budgets, incurred by different applications/devices, is essential for deploying deep models in real-world applications. To achieve this goal, existing methods often perform an independent architecture search process for each target budget, which is very inefficient yet unnecessary. More critically, these independent search processes cannot share their learned knowledge (i.e., the distribution of good architectures) with each other and thus often result in limited search results. To address these issues, we propose a Pareto-aware Neural Architecture Generator (PNAG) which only needs to be trained once and dynamically produces the Pareto optimal architecture for any given budget via inference. To train our PNAG, we learn the whole Pareto frontier by jointly finding multiple Pareto optimal architectures under diverse budgets. Such a joint search algorithm not only greatly reduces the overall search cost but also improves the search results. Extensive experiments on three hardware platforms (i.e., mobile device, CPU, and GPU) show the superiority of our method over existing methods.
Mixture-of-Expert (MoE) models have obtained state-of-the-art performance in Neural Machine Translation (NMT) tasks. Existing works in MoE mostly consider a homogeneous design where the same number of experts of the same size are placed uniformly throughout the network. Furthermore, existing MoE works do not consider computational constraints (e.g., FLOPs, latency) to guide their design. To this end, we develop AutoMoE -- a framework for designing heterogeneous MoE's under computational constraints. AutoMoE leverages Neural Architecture Search (NAS) to obtain efficient sparse MoE sub-transformers with 4x inference speedup (CPU) and FLOPs reduction over manually designed Transformers, with parity in BLEU score over dense Transformer and within 1 BLEU point of MoE SwitchTransformer, on aggregate over benchmark datasets for NMT. Heterogeneous search space with dense and sparsely activated Transformer modules (e.g., how many experts? where to place them? what should be their sizes?) allows for adaptive compute -- where different amounts of computations are used for different tokens in the input. Adaptivity comes naturally from routing decisions which send tokens to experts of different sizes. AutoMoE code, data, and trained models are available at https://aka.ms/AutoMoE.
Deep learning (DL) has been widely adopted those last years but they are computing-intensive method. Therefore, scientists proposed diverse optimization to accelerate their predictions for end-user applications. However, no single inference framework currently dominates in terms of performance. This paper takes a holistic approach to conduct an empirical comparison and analysis of four representative DL inference frameworks. First, given a selection of CPU-GPU configurations, we show that for a specific DL framework, different configurations of its settings may have a significant impact on the prediction speed, memory, and computing power. Second, to the best of our knowledge, this study is the first to identify the opportunities for accelerating the ensemble of co-localized models in the same GPU. This measurement study provides an in-depth empirical comparison and analysis of four representative DL frameworks and offers practical guidance for service providers to deploy and deliver DL predictions.
We present Rieoptax, an open source Python library for Riemannian optimization in JAX. We show that many differential geometric primitives, such as Riemannian exponential and logarithm maps, are usually faster in Rieoptax than existing frameworks in Python, both on CPU and GPU. We support various range of basic and advanced stochastic optimization solvers like Riemannian stochastic gradient, stochastic variance reduction, and adaptive gradient methods. A distinguishing feature of the proposed toolbox is that we also support differentially private optimization on Riemannian manifolds.