December 4, 2019

395 words 2 mins read

Using Siamese CNNs for removing duplicate entries from real estate listing databases

Using Siamese CNNs for removing duplicate entries from real estate listing databases

Aggregation of geospecific real estate databases results in duplicate entries for properties located near geographical boundaries. Sergey Ermolin and Olga Ermolin detail an approach for identifying duplicate entries via the analysis of images that accompany real estate listings that leverages a transfer learning Siamese architecture based on VGG-16 CNN topology.
Talk Title Using Siamese CNNs for removing duplicate entries from real estate listing databases
Speakers Sergey Ermolin (Intel), Olga Ermolin (MLS Listings)
Conference Strata Data Conference
Conf Tag Making Data Work
Location London, United Kingdom
Date May 22-24, 2018
URL Talk Page
Slides Talk Slides
Video

Real estate databases are geospecific (e.g., East Bay, North Bay, South Bay, etc). If a house to be put up for sale is located close to the geoboundary, a real estate listing agent will often list it in both databases. For example, a house located in Milpitas, CA, would often be listed in both East Bay and South Bay databases, although the content of both database entries may be different to appeal to the different demographics of each area. Real estate brokerage firms enter in cross-area sharing agreements, and there are efforts underway to create a nationwide sharing framework as well. Herein lies the problem: when data feeds from East Bay and South Bay databases are aggregated, this results in duplicate listings. Sergey Ermolin and Olga Ermolin detail an approach for identifying duplicate entries via the analysis of images that accompany real estate listings that leverages a transfer learning Siamese architecture based on VGG-16 CNN topology in TensorFlow 1.2. The curated dataset of over 3,000 images includes images of the front of the houses provided by MLS Listings, Inc., which contains entries (sets of JPEG images) that are a priori known to belong to duplicate real estate listings as well as those which are distinct. Before embarking on building a convolutional neural network, Sergey, Olga, and her team attempted a brute-force approach using a 1-nearest neighbor algorithm to establish a baseline for accuracy and precision of the prediction. Sergey and Olga explain why the brute-force nearest-neighbor approach was inadequate and how the CNN Siamese network was able to achieve accuracy of 69% and precision of 92%. To demonstrate that the implementation scales well with increased dataset, they also describe an implementation of the same CNN Siamese network in Spark’s BigDL framework and compare the results with those of TensorFlow.

prediction framework algorithm dataset spark tensorflow cnn database network broker transfer learning neural network
comments powered by Disqus
Automatic 3D MRI knee damage classification with 3D CNN using BigDL on Spark

Automatic 3D MRI knee damage classification with 3D CNN using BigDL on Spark

November 29, 2019

Damage to the meniscus is a physically limiting injury that can lead to further medical complications. Automatically classifying this damage at the time of an MRI scan would allow quicker and more accurate diagnosis. Jennie Wang, Valentina Pedoia, Berk Norman, and Yulia Tell offer an overview of their classification system built with 3D convolutional neural networks using BigDL on Apache Spark.
Deep learning for domain-specific entity extraction from unstructured text

Deep learning for domain-specific entity extraction from unstructured text

November 27, 2019

Mohamed AbdelHady and Zoran Dzunic demonstrate how to build a domain-specific entity extraction system from unstructured text using deep learning. In the model, domain-specific word embedding vectors are trained on a Spark cluster using millions of PubMed abstracts and then used as features to train a LSTM recurrent neural network for entity extraction.
Distributed clinical models: Inference without sharing patient data

Distributed clinical models: Inference without sharing patient data

November 26, 2019

Clinical collaboration benefits from pooling data to train models from large datasets, but it's hampered by concerns about sharing data. Balasubramanian Narasimhan, John-Mark Agosta, and Philip Lavori outline a privacy-preserving alternative that creates statistical models equivalent to one from the entire dataset.
Improving user-merchant propensity modeling using neural collaborative filtering and wide and deep models on Spark BigDL at scale

Improving user-merchant propensity modeling using neural collaborative filtering and wide and deep models on Spark BigDL at scale

November 24, 2019

Sergey Ermolin and Suqiang Song demonstrate how to use Spark BigDL wide and deep and neural collaborative filtering (NCF) algorithms to predict a users probability of shopping at a particular offer merchant during a campaign period. Along the way, they compare the deep learning results with those obtained by MLlibs alternating least squares (ALS) approach.
Not your parents' machine learning: How to ship an XGBoost churn prediction app in under four weeks

Not your parents' machine learning: How to ship an XGBoost churn prediction app in under four weeks

November 22, 2019

Machine learning is a pivotal technology. However, bringing an ML application to life often requires overcoming bottlenecks not just in the model code but in operationalizing the end-to-end system itself. Goodman Gu shares a case study from a leading SaaS company that quickly and easily built, trained, optimized, and deployed an XGBoost churn prediction ML app at scale with Amazon SageMaker.
Sprouted clams and stanky bean: When machine learning makes mistakes

Sprouted clams and stanky bean: When machine learning makes mistakes

November 20, 2019

At AIweirdness.com Janelle Shane posts the results of neural network experiments gone delightfully wrong. But machine learning mistakes can also be very embarrassing or even dangerous. Using silly datasets as examples, Janelle talks about some ways that algorithms fail.