No commercial reproduction, distribution, display or performance rights in this work are provided. Achille et al., "Task2Vec: Task Embedding for Meta-Learning," 2019 IEEE/CVF International Conference on Computer Vision (ICCV), Seoul, Korea (South), 2019, pp. Selecting a feature extractor with task embedding yields performance close to the best available feature extractor, with substantially less computational effort than exhaustively training and evaluating all available models.Ī. We present a simple meta-learning framework for learning a metric on embeddings that is capable of predicting which feature extractors will perform well on which task. We demonstrate the practical value of this framework for the meta-task of selecting a pre-trained feature extractor for a novel task. We demonstrate that this embedding is capable of predicting task similarities that match our intuition about semantic and taxonomic relations between different visual tasks. This provides a fixed-dimensional embedding of the task that is independent of details such as the number of classes and requires no understanding of the class label semantics. Given a dataset with ground-truth labels and a loss function, we process images through a "probe network" and compute an embedding based on estimates of the Fisher information matrix associated with the probe network parameters. ![]() Model trained on ImageNet.We introduce a method to generate vectorial representations of visual classification tasks which can be used to reason about the nature of those tasks and their relations. Outperforms the standard practice of fine-tuning a generic Select an expert from a collection of 156 feature extractors Our experiments show that using TASK2VEC to Ing set size, mimicking the heavy-tailed distribution of real. ![]() These tasks vary in the level ofĭifficulty and have orders of magnitude variation in train. Given a dataset with ground-truth labels and a loss function, we process images through a 'probe network' and compute an. Abstract: We introduce a method to generate vectorial representations of visual classification tasks which can be used to reason about the nature of those tasks and their relations. We present large-scale experiments on a library of 1,460įine-grained classification tasks constructed from existingĬomputer vision datasets. Task2Vec: Task Embedding for Meta-Learning. We formulate thisĪs a meta-learning problem where the objective is to findĪn embedding such that that models whose embeddings areĬlose to a task exhibit good performance on that task. Sign a joint embedding of models and tasks in the same vec. To select an appropriate pre-trained model, we de. Ticularly valuable when there is insufficient data to train orįine-tune a generic model, and transfer of knowledge is es. Ple, we study the problem of selecting the best pre-trainedįeature extractor to solve a new task (Sect. Our task embedding can be used to reason about the During my PhD I have also been a research scientist intern at Deep Mind and. I graduated in 2019 from the Computer Science Department of UCLA, working with Prof. Probe network are discriminative for the task (Sect. I am an Applied Scientist working in computer vision and deep learning at Amazon AI (Pasadena) and Caltech (visiting scholar). Of the input domain, and which features extracted by the Multaneously encodes the “difficulty” of the task, statistics Representation of the task which is independent of, e.g., how Network are fixed, the FIM provides a fixed-dimensional Thus, our method is more practical than TASK2VEC 31. ![]() While TASK2VEC 31 needs data and label in target domains. Since the architecture and weights of the probe The advantage of our method compared with TASK2VEC 31 is that our method does not require any data in target domains, i.e., training of our method is only performed on source domains. Work filter parameters to capture the structure of the task The diagonal Fisher Information Matrix (FIM) of the net. Ral network which we call a “probe network”, and compute The data through a pre-trained reference convolutional neu. T ASK 2V EC : Task Embedding for Meta-Learning Alessandro Achille 1, 2 Michael Lam 1 Rahul Tewari 1 Avinash Ravichandran 1 Subhransu Maji 1, 3 Charless Fowlkes 1, 4 Stefano Soatto 1, 2 Pietro Perona 1, 5 Ni=1 of labeled samples, we feed
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |