Abstract Representations of the world environment play a crucial role in artificial intelligence. It is often inefficient to conduct reasoning and inference directly in the space of raw sensory representations, such as pixel values of images. Representation learning allows us to automatically discover suitable representations from raw sensory data. For example, given raw sensory data, a deep neural network learns nonlinear representations at its hidden layers, which are subsequently used for classification (or regression) at its output layer. This happens implicitly during training through minimizing a supervised or unsupervised loss. In this letter, we study the dynamics of such implicit nonlinear representation learning. We identify a pair of a new assumption and a novel condition, called the on-model structure assumption and the data architecture alignment condition. Under the on-model structure assumption, the data architecture alignment condition is shown to be sufficient for the global convergence and necessary for global optimality. Moreover, our theory explains how and when increasing network size does and does not improve the training behaviors in the practical regime. Our results provide practical guidance for designing a model structure; for example, the on-model structure assumption can be used as a justification for using a particular model structure instead of others. As an application, we then derive a new training framework, which satisfies the data architecture alignment condition without assuming it by automatically modifying any given training algorithm dependent on data and architecture. Given a standard training algorithm, the framework running its modified version is empirically shown to maintain competitive (practical) test performances while providing global convergence guarantees for deep residual neural networks with convolutions, skip connections, and batch normalization with standard benchmark data sets, including MNIST, CIFAR-10, CIFAR-100, Semeion, KMNIST, and SVHN.
more »
« less
This content will become publicly available on September 15, 2024
“No Free Lunch” in Neural Architectures? A Joint Analysis of Expressivity, Convergence, and Generalization
The prosperity of deep learning and automated machine learning (AutoML) is largely rooted in the development of novel neural networks -- but what defines and controls the "goodness" of networks in an architecture space? Test accuracy, a golden standard in AutoML, is closely related to three aspects: (1) expressivity (how complicated functions a network can approximate over the training data); (2) convergence (how fast the network can reach low training error under gradient descent); (3) generalization (whether a trained network can be generalized from the training data to unseen samples with low test error). However, most previous theory papers focus on fixed model structures, largely ignoring sophisticated networks used in practice. To facilitate the interpretation and understanding of the architecture design by AutoML, we target connecting a bigger picture: how does the architecture jointly impact its expressivity, convergence, and generalization? We demonstrate the "no free lunch" behavior in networks from an architecture space: given a fixed budget on the number of parameters, there does not exist a single architecture that is optimal in all three aspects. In other words, separately optimizing expressivity, convergence, and generalization will achieve different networks in the architecture space. Our analysis can explain a wide range of observations in AutoML. Experiments on popular benchmarks confirm our theoretical analysis. Our codes are attached in the supplement.
more »
« less
- Award ID(s):
- 2133861
- NSF-PAR ID:
- 10480848
- Publisher / Repository:
- AutoML Conference 2023
- Date Published:
- Journal Name:
- AutoML Conference 2023
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Despite the phenomenal success of deep neural networks in a broad range of learning tasks, there is a lack of theory to understand the way they work. In particular, Convolutional Neural Networks (CNNs) are known to perform much better than Fully-Connected Networks (FCNs) on spatially structured data: the architectural structure of CNNs benefits from prior knowledge on the features of the data, for instance their translation invariance. The aim of this work is to understand this fact through the lens of dynamics in the loss landscape. We introduce a method that maps a CNN to its equivalent FCN (denoted as eFCN). Such an embedding enables the comparison of CNN and FCN training dynamics directly in the FCN space. We use this method to test a new training protocol, which consists in training a CNN, embedding it to FCN space at a certain relax time'', then resuming the training in FCN space. We observe that for all relax times, the deviation from the CNN subspace is small, and the final performance reached by the eFCN is higher than that reachable by a standard FCN of same architecture. More surprisingly, for some intermediate relax times, the eFCN outperforms the CNN it stemmed, by combining the prior information of the CNN and the expressivity of the FCN in a complementary way. The practical interest of our protocol is limited by the very large size of the highly sparse eFCN. However, it offers interesting insights into the persistence of architectural bias under stochastic gradient dynamics. It shows the existence of some rare basins in the FCN loss landscape associated with very good generalization. These can only be accessed thanks to the CNN prior, which helps navigate the landscape during the early stages of optimization.more » « less
-
null (Ed.)Artificial neural networks (NNs) in deep learning systems are critical drivers of emerging technologies such as computer vision, text classification, and natural language processing. Fundamental to their success is the development of accurate and efficient NN models. In this article, we report our work on Deep-n-Cheap—an open-source automated machine learning (AutoML) search framework for deep learning models. The search includes both architecture and training hyperparameters and supports convolutional neural networks and multi-layer perceptrons, applicable to multiple domains. Our framework is targeted for deployment on both benchmark and custom datasets, and as a result, offers a greater degree of search space customizability as compared to a more limited search over only pre-existing models from literature. We also introduce the technique of ‘search transfer’, which demonstrates the generalization capabilities of the models found by our framework to multiple datasets. Deep-n-Cheap includes a user-customizable complexity penalty which trades off performance with training time or number of parameters. Specifically, our framework can find models with performance comparable to state-of-the- art while taking 1–2 orders of magnitude less time to train than models from other AutoML and model search frameworks. Additionally, we investigate and develop insight into the search process that should aid future development of deep learning models.more » « less
-
Obeid, I. ; Selesnik, I. ; Picone, J. (Ed.)The Neuronix high-performance computing cluster allows us to conduct extensive machine learning experiments on big data [1]. This heterogeneous cluster uses innovative scheduling technology, Slurm [2], that manages a network of CPUs and graphics processing units (GPUs). The GPU farm consists of a variety of processors ranging from low-end consumer grade devices such as the Nvidia GTX 970 to higher-end devices such as the GeForce RTX 2080. These GPUs are essential to our research since they allow extremely compute-intensive deep learning tasks to be executed on massive data resources such as the TUH EEG Corpus [2]. We use TensorFlow [3] as the core machine learning library for our deep learning systems, and routinely employ multiple GPUs to accelerate the training process. Reproducible results are essential to machine learning research. Reproducibility in this context means the ability to replicate an existing experiment – performance metrics such as error rates should be identical and floating-point calculations should match closely. Three examples of ways we typically expect an experiment to be replicable are: (1) The same job run on the same processor should produce the same results each time it is run. (2) A job run on a CPU and GPU should produce identical results. (3) A job should produce comparable results if the data is presented in a different order. System optimization requires an ability to directly compare error rates for algorithms evaluated under comparable operating conditions. However, it is a difficult task to exactly reproduce the results for large, complex deep learning systems that often require more than a trillion calculations per experiment [5]. This is a fairly well-known issue and one we will explore in this poster. Researchers must be able to replicate results on a specific data set to establish the integrity of an implementation. They can then use that implementation as a baseline for comparison purposes. A lack of reproducibility makes it very difficult to debug algorithms and validate changes to the system. Equally important, since many results in deep learning research are dependent on the order in which the system is exposed to the data, the specific processors used, and even the order in which those processors are accessed, it becomes a challenging problem to compare two algorithms since each system must be individually optimized for a specific data set or processor. This is extremely time-consuming for algorithm research in which a single run often taxes a computing environment to its limits. Well-known techniques such as cross-validation [5,6] can be used to mitigate these effects, but this is also computationally expensive. These issues are further compounded by the fact that most deep learning algorithms are susceptible to the way computational noise propagates through the system. GPUs are particularly notorious for this because, in a clustered environment, it becomes more difficult to control which processors are used at various points in time. Another equally frustrating issue is that upgrades to the deep learning package, such as the transition from TensorFlow v1.9 to v1.13, can also result in large fluctuations in error rates when re-running the same experiment. Since TensorFlow is constantly updating functions to support GPU use, maintaining an historical archive of experimental results that can be used to calibrate algorithm research is quite a challenge. This makes it very difficult to optimize the system or select the best configurations. The overall impact of all of these issues described above is significant as error rates can fluctuate by as much as 25% due to these types of computational issues. Cross-validation is one technique used to mitigate this, but that is expensive since you need to do multiple runs over the data, which further taxes a computing infrastructure already running at max capacity. GPUs are preferred when training a large network since these systems train at least two orders of magnitude faster than CPUs [7]. Large-scale experiments are simply not feasible without using GPUs. However, there is a tradeoff to gain this performance. Since all our GPUs use the NVIDIA CUDA® Deep Neural Network library (cuDNN) [8], a GPU-accelerated library of primitives for deep neural networks, it adds an element of randomness into the experiment. When a GPU is used to train a network in TensorFlow, it automatically searches for a cuDNN implementation. NVIDIA’s cuDNN implementation provides algorithms that increase the performance and help the model train quicker, but they are non-deterministic algorithms [9,10]. Since our networks have many complex layers, there is no easy way to avoid this randomness. Instead of comparing each epoch, we compare the average performance of the experiment because it gives us a hint of how our model is performing per experiment, and if the changes we make are efficient. In this poster, we will discuss a variety of issues related to reproducibility and introduce ways we mitigate these effects. For example, TensorFlow uses a random number generator (RNG) which is not seeded by default. TensorFlow determines the initialization point and how certain functions execute using the RNG. The solution for this is seeding all the necessary components before training the model. This forces TensorFlow to use the same initialization point and sets how certain layers work (e.g., dropout layers). However, seeding all the RNGs will not guarantee a controlled experiment. Other variables can affect the outcome of the experiment such as training using GPUs, allowing multi-threading on CPUs, using certain layers, etc. To mitigate our problems with reproducibility, we first make sure that the data is processed in the same order during training. Therefore, we save the data from the last experiment and to make sure the newer experiment follows the same order. If we allow the data to be shuffled, it can affect the performance due to how the model was exposed to the data. We also specify the float data type to be 32-bit since Python defaults to 64-bit. We try to avoid using 64-bit precision because the numbers produced by a GPU can vary significantly depending on the GPU architecture [11-13]. Controlling precision somewhat reduces differences due to computational noise even though technically it increases the amount of computational noise. We are currently developing more advanced techniques for preserving the efficiency of our training process while also maintaining the ability to reproduce models. In our poster presentation we will demonstrate these issues using some novel visualization tools, present several examples of the extent to which these issues influence research results on electroencephalography (EEG) and digital pathology experiments and introduce new ways to manage such computational issues.more » « less
-
Abstract. Advances in ambient environmental monitoring technologies are enabling concerned communities and citizens to collect data to better understand their local environment and potential exposures. These mobile, low-cost tools make it possible to collect data with increased temporal and spatial resolution, providing data on a large scale with unprecedented levels of detail. This type of data has the potential to empower people to make personal decisions about their exposure and support the development of local strategies for reducing pollution and improving health outcomes. However, calibration of these low-cost instruments has been a challenge. Often, a sensor package is calibrated via field calibration. This involves colocating the sensor package with a high-quality reference instrument for an extended period and then applying machine learning or other model fitting technique such as multiple linear regression to develop a calibration model for converting raw sensor signals to pollutant concentrations. Although this method helps to correct for the effects of ambient conditions (e.g., temperature) and cross sensitivities with nontarget pollutants, there is a growing body of evidence that calibration models can overfit to a given location or set of environmental conditions on account of the incidental correlation between pollutant levels and environmental conditions, including diurnal cycles. As a result, a sensor package trained at a field site may provide less reliable data when moved, or transferred, to a different location. This is a potential concern for applications seeking to perform monitoring away from regulatory monitoring sites, such as personal mobile monitoring or high-resolution monitoring of a neighborhood. We performed experiments confirming that transferability is indeed a problem and show that it can be improved by collecting data from multiple regulatory sites and building a calibration model that leverages data from a more diverse data set. We deployed three sensor packages to each of three sites with reference monitors (nine packages total) and then rotated the sensor packages through the sites over time. Two sites were in San Diego, CA, with a third outside of Bakersfield, CA, offering varying environmental conditions, general air quality composition, and pollutant concentrations. When compared to prior single-site calibration, the multisite approach exhibits better model transferability for a range of modeling approaches. Our experiments also reveal that random forest is especially prone to overfitting and confirm prior results that transfer is a significant source of both bias and standard error. Linear regression, on the other hand, although it exhibits relatively high error, does not degrade much in transfer. Bias dominated in our experiments, suggesting that transferability might be easily increased by detecting and correcting for bias. Also, given that many monitoring applications involve the deployment of many sensor packages based on the same sensing technology, there is an opportunity to leverage the availability of multiple sensors at multiple sites during calibration to lower the cost of training and better tolerate transfer. We contribute a new neural network architecture model termed split-NN that splits the model into two stages, in which the first stage corrects for sensor-to-sensor variation and the second stage uses the combined data of all the sensors to build a model for a single sensor package. The split-NN modeling approach outperforms multiple linear regression, traditional two- and four-layer neural networks, and random forest models. Depending on the training configuration, compared to random forest the split-NN method reduced error 0 %–11 % for NO2 and 6 %–13 % for O3.more » « less
