Revisiting Unsupervised Meta-Learning via the Characteristics of Few-Shot Tasks

The code repository for "Revisiting Unsupervised Meta-Learning via the Characteristics of Few-Shot Tasks"

Overview

We first analyze the factors to meta-train a UML method and propose SES and SNS as two key ingredients towards a strong UML baseline. Then, we propose HMS and TSP-Head to further utilize the characteristic of tasks from different aspects, which additionally improve either lower or higher shots scenarios.

Unsupervised Meta Learning Results

Experimental results on few-shot learning datasets with ResNet-12 backbone (Same as this repo). We report average results with 10,000 randomly sampled few-shot learning episodes for stablized evaluation.

MiniImageNet Dataset with ConvNet

MiniImageNet Dataset with ResNet-12

Prerequisites

The following packages are required to run the scripts:

• PyTorch and torchvision

• Package tensorboardX

• Dataset: please download the dataset and put images into the folder data/[name of the dataset, miniimagenet or cub]/images

Dataset

MiniImageNet Dataset

The MiniImageNet dataset is a subset of the ImageNet that includes a total number of 100 classes and 600 examples per class. We follow the previous setup, and use 64 classes as SEEN categories, 16 and 20 as two sets of UNSEEN categories for model validation and evaluation, respectively.

CUB Dataset

Caltech-UCSD Birds (CUB) 200-2011 dataset is initially designed for fine-grained classification. It contains in total 11,788 images of birds over 200 species. On CUB, we randomly sampled 100 species as SEEN classes, and another two 50 species are used as two UNSEEN sets. We crop all images with given bounding boxes before training. We only test CUB with the ConvNet backbone in our work.

TieredImageNet Dataset

TieredImageNet is a large-scale dataset with more categories, which contains 351, 97, and 160 categoriesfor model training, validation, and evaluation, respectively. The dataset can also be download from here. We only test TieredImageNet with ResNet backbone in our work.

Check this for details of data downloading and preprocessing.

Code Structures

To reproduce our experiments with FEAT, please use train_fsl.py. There are four parts in the code.

• model: It contains the main files of the code, including the few-shot learning trainer, the dataloader, the network architectures, and baseline and comparison models.
• data: Images and splits for the data sets.
• saves: The pre-trained weights of different networks.
• checkpoints: To save the trained models.
• script: Example commands.

Model Training and Evaluation

Please use train.py and follow the instructions below. FEAT meta-learns the embedding adaptation process such that all the training instance embeddings in a task is adapted, based on their contextual task information, using Transformer. The file will automatically evaluate the model on the meta-test set with 10,000 tasks after given epochs.

Arguments

The train.py takes the following command line options (details are in the model/utils.py):

Task Related Arguments

• dataset: Option for the dataset (MiniImageNet, TieredImageNet,CIFAR-FS , FC100 , or CUB), default to MiniImageNet

• data_root: root directory for the dataset.

• way: The number of classes in a few-shot task during meta-training, default to 5

• eval_way: The number of classes in a few-shot task during meta-test, default to 5

• shot: Number of instances in each class in a few-shot task during meta-training, default to 1

• eval_shot: Number of instances in each class in a few-shot task during meta-test, default to 1

• query: Number of instances in each class to evaluate the performance during meta-training, default to 15

• eval_query: Number of instances in each class to evaluate the performance during meta-test, default to 15

Optimization Related Arguments

• max_epoch: The maximum number of training epochs, default to 200

• episodes_per_epoch: The number of tasks sampled in each epoch, default to 100

• num_eval_episodes: The number of tasks sampled from the meta-val set to evaluate the performance of the model (note that we fix sampling 10,000 tasks from the meta-test set during final evaluation), default to 200

• lr: Learning rate for the model, default to 0.0001 with pre-trained weights

• lr_scheduler: The scheduler to set the learning rate (step, multistep, or cosine), default to step

• step_size: The step scheduler to decrease the learning rate. Set it to a single value if choose the step scheduler and provide multiple values when choosing the multistep scheduler. Default to 20

• gamma: Learning rate ratio for step or multistep scheduler, default to 0.2

• augment: Whether to do data augmentation or not during meta-training, default to False

• mom: The momentum value for the SGD optimizer, default to 0.9

• weight_decay: The weight_decay value for SGD optimizer, default to 0.0005

Model Related Arguments

• model_class: The model to use. We provide implementations for uml baseline (ProtoNet) and TSP-Head. (Note that we implement HMS as an additional component, see additional.) Default to ProtoNet

• similarity: Similarity measure of two instances. We provide four implementation euclidean ,inner product (dot), cosine and sns . Default to sns

• backbone_class: Types of the encoder, i.e., the convolution network (ConvNet), ResNet-12 (Res12), or Wide ResNet (WRN), default to ConvNet

• temperature: Temperature over the logits, we #divide# logits with this value. It is useful when meta-learning with pre-trained weights. Default to 1

• temperature2: Temperature over the logits in the regularizer, we divide logits with this value. This is specially designed for the contrastive regularizer. Default to 1

• additional: Additional component that can add to model. Can be HMS or none. Default to none.

Arguments for TSP-Head:

• lr_mul: This is specially designed TSP-Head. The learning rate for the top layer will be multiplied by this value (usually with faster learning rate). Default to 10

• t_heads: Number of heads for Multi-Head Transformer. Default to 8.

• t_layers: Number of layers for Transformer. Default to 1.

• t_dropout: Drop out rate for Transformer. Default to 0.2.

• t_dim: Dimension of Keys and Values for Transformer. -1 means equaling to dimension of input feature vector. Default to -1.

Arguments for HMS:

• hard_negs : Number of hard negative samples for each query instance. Default to 10.

• strength : Mix up strength. The mix up factor will be draw from U(0,strength). Default to 0.5.

Other Arguments

• orig_imsize: Whether to resize the images before loading the data into the memory. -1 means we do not resize the images and do not read all images into the memory. Default to -1

• multi_gpu: Whether to use multiple gpus during meta-training, default to False

• gpu: The index of GPU to use. Please provide multiple indexes if choose multi_gpu. Default to 0

• log_interval: How often to log the meta-training information, default to every 50 tasks

• eval_interval: How often to validate the model over the meta-val set, default to every 1 epoch

• save_dir: The path to save the learned models, default to ./checkpoints

Running the command without arguments will train the models with the default hyper-parameter values. Loss changes will be recorded as a tensorboard file.

Training scripts

For example, to train UML baseline with ResNet-12 backbone on MiniImageNet:

$ python train.py --eval_all --unsupervised --batch_size 32 --augment 'AMDIM' --num_tasks 256 --max_epoch 200 --model_class ProtoNet --backbone_class Res12 --dataset MiniImageNet --way 5 --shot 1 --query 5 --eval_query 15 --temperature 1 --temperature2 1 --lr 0.03 --lr_scheduler cosine --gpu 0 --eval_interval 2 --similarity sns

to train TSP-Head with ResNet-12 backbone on MiniImageNet

$ python train.py --eval_all --unsupervised --batch_size 32 --augment 'AMDIM' --num_tasks 256 --max_epoch 200 --model_class TSPHead --backbone_class Res12 --dataset MiniImageNet --way 5 --shot 1 --query 5 --eval_query 15 --temperature 1 --temperature2 1 --lr 0.03 --lr_mul 10 --lr_scheduler cosine --gpu 0 --eval_interval 2 --similarity sns --t_heads 8

to train HMS with ResNet-12 backbone on MiniImageNet

$ python train.py --eval_all --unsupervised --batch_size 32 --augment 'AMDIM' --num_tasks 256 --max_epoch 200 --model_class ProtoNet --backbone_class Res12 --dataset MiniImageNet --way 5 --shot 1 --query 5 --eval_query 15 --balance 0 --temperature 1 --temperature2 1 --lr 0.03 --lr_mul 1 --lr_scheduler cosine --gpu 0 --eval_interval 2 --similarity sns --additional HMS --strength 0.5

Acknowledgment

We thank the following repos providing helpful components/functions in our work.

• ProtoNet

• MatchingNet

• PFA

• Transformer

• MetaOptNet

• FEAT