An official website of the United States government
Ground truth depth information is necessary for many computer vision tasks. Collecting this information is chal-lenging, especially for outdoor scenes. In this work, we propose utilizing single-view depth prediction neural networks pre-trained on synthetic scenes to generate relative depth, which we call pseudo-depth. This approach is a less expen-sive option as the pre-trained neural network obtains ac-curate depth information from synthetic scenes, which does not require any expensive sensor equipment and takes less time. We measure the usefulness of pseudo-depth from pre-trained neural networks by training indoor/outdoor binary classifiers with and without it. We also compare the difference in accuracy between using pseudo-depth and ground truth depth. We experimentally show that adding pseudo-depth to training achieves a 4.4% performance boost over the non-depth baseline model on DIODE, a large stan-dard test dataset, retaining 63.8% of the performance boost achieved from training a classifier on RGB and ground truth depth. It also boosts performance by 1.3% on another dataset, SUN397, for which ground truth depth is not avail-able. Our result shows that it is possible to take information obtained from a model pre-trained on synthetic scenes and successfully apply it beyond the synthetic domain to real-world data.
more »
« less
Injecting Domain Knowledge from Empirical Interatomic Potentials to Neural Networks for Predicting Material Properties
For decades, atomistic modeling has played a crucial role in predicting the behavior of materials in numerous fields ranging from nanotechnology to drug discovery. The most accurate methods in this domain are rooted in first-principles quantum mechanical calculations such as density functional theory (DFT). Because these methods have remained computationally prohibitive, practitioners have traditionally focused on defining physically motivated closed-form expressions known as empirical interatomic potentials (EIPs) that approximately model the interactions between atoms in materials. In recent years, neural network (NN)-based potentials trained on quantum mechanical (DFT-labeled) data have emerged as a more accurate alternative to conventional EIPs. However, the generalizability of these models relies heavily on the amount of labeled training data, which is often still insufficient to generate models suitable for general-purpose applications. In this paper, we propose two generic strategies that take advantage of unlabeled training instances to inject domain knowledge from conventional EIPs to NNs in order to increase their generalizability. The first strategy, based on weakly supervised learning, trains an auxiliary classifier on EIPs and selects the best-performing EIP to generate energies to supplement the ground-truth DFT energies in training the NN. The second strategy, based on transfer learning, first pretrains the NN on a large set of easily obtainable EIP energies, and then fine-tunes it on ground-truth DFT energies. Experimental results on three benchmark datasets demonstrate that the first strategy improves baseline NN performance by 5% to 51% while the second improves baseline performance by up to 55%. Combining them further boosts performance.
more »
« less
- Award ID(s):
- 1834251
- PAR ID:
- 10382006
- Date Published:
- Journal Name:
- NeurIPS 2022
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Key Points A framework merging unsupervised clustering and supervised convolutional neural network (CNN) for lightning classification is developed Clustering of positive polarity energetic lightning radio pulses (>150 kA) identifies three processes: +EIPs (6%–7%), +NBEs, and +CGs CNNs detect 95.2% of manually identified +EIPs with up to 98.7% accuracy, enabling studying EIP‐TGF link with lower peak current (>50 kA)more » « less
-
Neural Network potentials are developed which accurately make and break bonds for use in molecular simulations. We report a neural network potential that can describe the potential energy surface for carbon–carbon bond dissociation with less than 1 kcal mol−1 error compared to complete active space second-order perturbation theory (CASPT2), and maintains this accuracy for both the minimum energy path and molecular dynamic calculations up to 2000 K. We utilize a transfer learning algorithm to develop neural network potentials to generate potential energy surfaces; this method aims to use the minimum amount of CASPT2 data on small systems to train neural network potentials while maintaining excellent transferability to larger systems. First, we generate homolytic carbon–carbon bond dissociation data of small size alkanes with density functional theory (DFT) energies to train the potentials to accurately predict bond dissociation at the DFT level. Then, using transfer learning, we retrained the neural network potential to the CASPT2 level of accuracy. We demonstrate that the neural network potential only requires bond dissociation data of a few small alkanes to accurately predict bond dissociation energy in larger alkanes. We then perform additional training on molecular dynamic simulations to refine our neural network potentials to obtain high accuracy for general use in molecular simulation. This training algorithm is generally applicable to any type of bond or any level of theory and will be useful for the generation of new reactive neural network potentials.more » « less
-
Obtaining annotations for large training sets is expen- sive, especially in settings where domain knowledge is re- quired, such as behavior analysis. Weak supervision has been studied to reduce annotation costs by using weak la- bels from task-specific labeling functions (LFs) to augment ground truth labels. However, domain experts still need to hand-craft different LFs for different tasks, limiting scal- ability. To reduce expert effort, we present AutoSWAP: a framework for automatically synthesizing data-efficient task-level LFs. The key to our approach is to efficiently represent expert knowledge in a reusable domain-specific language and more general domain-level LFs, with which we use state-of-the-art program synthesis techniques and a small labeled dataset to generate task-level LFs. Addition- ally, we propose a novel structural diversity cost that allows for efficient synthesis of diverse sets of LFs, further improv- ing AutoSWAP’s performance. We evaluate AutoSWAP in three behavior analysis domains and demonstrate that Au- toSWAP outperforms existing approaches using only a frac- tion of the data. Our results suggest that AutoSWAP is an effective way to automatically generate LFs that can signif- icantly reduce expert effort for behavior analysis.more » « less
-
Recurrent Classifier Chains (RCCs) are a leading approach for multi-label classification as they directly model the interdependencies between classes. Unfortunately, existing RCCs assume that every training instance is completely labeled with all its ground truth classes. In practice often only a subset of an instance's labels are annotated, while the annotations for other classes are missing. RCCs fail in this missing label scenario, predicting many false negatives and potentially missing important classes. In this work, we propose Robust-RCC, the first strategy for tackling this open problem of RCCs failing for multi-label missing-label data. Robust-RCC is a new type of deep recurrent classifier chain empowered to model inter-class relationships essential for predicting the complete label set most likely to match the ground truth. The key to Robust-RCC is the design of the Multi Incomplete Label Risk (MILR) function, which we prove to be equal in expectation to the true risk of the ground truth full label set despite being computed from incompletely labeled data. Our experimental study demonstrates that Robust-RCC consistently beats six state-of-of-the-art methods by as much as 30% in predicting the true labels.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- The DOI is not currently available.
