Learning Recurrent Representations for Hierarchical Behavior Modeling
Authors: Eyrun Eyjolfsdottir, Kristin Branson, Yisong Yue, Pietro Perona
ICLR 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We test our framework on two types of tracking data, fruit fly behavior and online handwriting. Our results show that 1) taking advantage of unlabeled sequences, by predicting future motion, significantly improves action detection performance when training labels are scarce, 2) the network learns to represent high level phenomena such as writer identity and fly gender, without supervision, and 3) simulated motion trajectories, generated by treating motion prediction as input to the network, look realistic and may be used to qualitatively evaluate whether the model has learnt generative control rules. |
| Researcher Affiliation | Academia | Eyrun Eyjolfsdottir1, Kristin Branson2, Yisong Yue1, & Pietro Perona1 1California Institute of Technology, 2Janelia Research Campus HHMI |
| Pseudocode | No | The paper describes the model architecture and equations but does not include any structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | Further training details are discussed in supplementary material1. 1www.vision.caltech.edu/ eeyjolfs/behavior_modeling |
| Open Datasets | Yes | Fly-vs-fly (Eyjolfsdottir et al., 2014) contains pairs of fruit flies... IAM-On DB (Liwicki & Bunke, 2005) contains handwritten text... All data, along with details about training and test splits, will be available in supplementary material. |
| Dataset Splits | No | The paper states, "All data, along with details about training and test splits, will be available in supplementary material," but does not explicitly provide the specific percentages or counts for training/validation/test splits within the main text. |
| Hardware Specification | No | The paper does not provide specific details about the hardware (e.g., GPU/CPU models, memory) used for running the experiments. |
| Software Dependencies | No | The paper states "Our model is implemented in Tensorflow (Abadi et al., 2015)," but does not provide specific version numbers for TensorFlow or any other software dependencies. |
| Experiment Setup | Yes | Model details: We trained a separate model for each dataset, using a sequence length of 50, a batch size of 20, and 51 bins per dimension for motion prediction. For fly behavior data we used 2 levels of GRU cells (4 cells total) of 100 units each, and for handwriting we used 3 levels of GRU cells (6 cells total) of 200 units each. |