Binary embeddings with structured hashed projections
Authors: Anna Choromanska, Krzysztof Choromanski, Mariusz Bojarski, Tony Jebara, Sanjiv Kumar, Yann LeCun
ICML 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We empirically verify our theoretical findings and show the dependence of learning via structured hashed projections on the performance of neural network as well as nearest neighbor classifier. This article is organized as follows: Section 5 shows experimental results |
| Researcher Affiliation | Collaboration | Anna Choromanska1 ACHOROMA@CIMS.NYU.EDU Krzysztof Choromanski1 KCHORO@GOOGLE.COM Mariusz Bojarski MBOJARSKI@NVIDIA.COM Tony Jebara JEBARA@CS.COLUMBIA.EDU Sanjiv Kumar SANJIVK@GOOGLE.COM Yann Le Cun YANN@CS.NYU.EDU |
| Pseudocode | No | The paper does not contain structured pseudocode or algorithm blocks. Section 3 describes the hashing mechanism with mathematical notation and a flow diagram, but not a formal algorithm block. |
| Open Source Code | No | The paper does not provide concrete access to source code for the methodology described. It states "All codes were implemented in Torch7" but provides no links or explicit statements about code availability. |
| Open Datasets | Yes | We performed experiments on MNIST dataset downloaded from http://yann.lecun.com/exdb/mnist/. |
| Dataset Splits | No | The paper mentions using the MNIST dataset and preprocessing it, but it does not provide specific dataset split information (exact percentages, sample counts, or citations to predefined splits) needed to reproduce the data partitioning into training, validation, and test sets. |
| Hardware Specification | No | The paper does not provide specific hardware details (exact GPU/CPU models, processor types with speeds, memory amounts, or detailed computer specifications) used for running its experiments. Table 1 refers to theoretical memory complexity of matrices, not experimental hardware. |
| Software Dependencies | No | All codes were implemented in Torch7. The paper mentions Torch7 but does not provide a specific version number for this or any other software dependency. |
| Experiment Setup | Yes | We first considered a simple model of the fully-connected feed-forward neural network with two hidden layers, where the first hidden layer had k units that use sign non-linearity (we explored k = {16, 32, 64, 128, 256, 512, 1024}), and the second hidden layer had 100 units that use Re LU non-linearity. Each experiment was initialized from a random set of parameters sampled uniformly within the unit cube, and was repeated 1000 times. All networks were trained for 30 epochs using SGD (Bottou, 1998). The learning rate was chosen from the set {0.0005, 0.001, 0.002, 0.005, 0.01, 0.02, 0.05, 0.1, 0.2, 0.5, 1} to minimize the test error. |