Deep Neural Decision Forests
Authors: Peter Kontschieder, Madalina Fiterau, Antonio Criminisi, Samuel Rota Bulò
IJCAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our approach compares favourably to other state-of-the-art deep models on a large-scale image classification task like Image Net...We show the efficacy of our approach on the challenging Image Net dataset for largescale image classification, where we obtain state-of-the-art results with no data augmentation. |
| Researcher Affiliation | Collaboration | Peter Kontschieder , Madalina Fiterau , Antonio Criminisi , Samuel Rota Bul o? Microsoft Research Cambridge, UK Stanford University California, USA ? Fondazione Bruno Kessler Trentino, Italy |
| Pseudocode | No | The paper describes the learning algorithms, such as the two-step optimization strategy and stochastic gradient descent, in paragraph text rather than structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide any statements about releasing open-source code or a link to a code repository for the described methodology. |
| Open Datasets | Yes | Image Net [Russakovsky et al., 2014] is a benchmark for large-scale image recognition tasks and its images are assigned to one out of 1000 possible ground truth labels. |
| Dataset Splits | Yes | The dataset contains 1.2M training images, 50.000 validation images and 100.000 test images with average dimensionality of 482x415 pixels. |
| Hardware Specification | Yes | We trained the network for 1000 epochs using (mini-) batches composed of 100.000 images (which was feasible due to distribution of the computational load to a cluster of 52 CPUs and 12 hosts, where each host is equipped with a NVIDIA Tesla K40 GPU). |
| Software Dependencies | No | The paper does not specify versions for any software dependencies, libraries, or frameworks used in the experiments. |
| Experiment Setup | Yes | We trained the network for 1000 epochs using (mini-) batches composed of 100.000 images... Following the implementation guideline for decision nodes in Section 4, we randomly selected 500 output dimensions of the respectively preceding layers in Goog Le Net for each decision function fn. |