Few-Shot Adversarial Domain Adaptation
Authors: Saeid Motiian, Quinn Jones, Seyed Iranmanesh, Gianfranco Doretto
NeurIPS 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We then extensively compare this approach to the state of the art in domain adaptation in two experiments: one using datasets for handwritten digit recognition, and one using datasets for visual object recognition. Table 1: MNIST-USPS-SVHN datasets. Classification accuracy for domain adaptation over the MNIST, USPS, and SVHN datasets. Table 2: Office dataset. Classification accuracy for domain adaptation over the 31 categories of the Office dataset. |
| Researcher Affiliation | Academia | Saeid Motiian, Quinn Jones, Seyed Mehdi Iranmanesh, Gianfranco Doretto Lane Department of Computer Science and Electrical Engineering West Virginia University {samotiian, qjones1, seiranmanesh, gidoretto}@mix.wvu.edu |
| Pseudocode | Yes | Algorithm 1 FADA algorithm |
| Open Source Code | No | The paper does not provide any statement about making its source code openly available or provide a link to a repository. |
| Open Datasets | Yes | We present results using the Office dataset [47], the MNIST dataset [32], the USPS dataset [24], and the SVHN dataset [40]. |
| Dataset Splits | No | We randomly selected n labeled samples per class from target domain data and used them in training. For the source domain, 20 examples per category for the Amazon domain, and 8 examples per category for the DSLR and Webcam domains are randomly selected for training for each split. Also, 3 labeled examples are randomly selected for each category in the target domain for training for each split. The rest of the target samples are used for testing. The paper describes training data and testing data but does not explicitly mention a separate validation split or how it's used if it exists. |
| Hardware Specification | No | The paper does not provide specific details about the hardware used for running the experiments. |
| Software Dependencies | No | Training for each stage was done using the Adam Optimizer [26]. The paper mentions general software like "Adam Optimizer" and "VGG-16 architecture" but does not provide specific version numbers for any software dependencies or libraries. |
| Experiment Setup | Yes | Similar to [32], we used 2 convolutional layers with 6 and 16 filters of 5 5 kernels followed by max-pooling layers and 2 fully connected layers with size 120 and 84 as the inference function g, and one fully connected layer with softmax activation as the prediction function h. Also, we used 2 fully connected layers with size 64 and 4 as DCD (4 groups classifier). Training for each stage was done using the Adam Optimizer [26]. For the inference function g, we used the convolutional layers of the VGG-16 architecture [53] followed by 2 fully connected layers with output size of 1024 and 128, respectively. For the prediction function h, we used a fully connected layer with softmax activation. Similar to [58], we used the weights pre-trained on the Image Net dataset [46] for the convolutional layers, and initialized the fully connected layers using all the source domain data. |