Presented By O'Reilly and Cloudera
Make Data Work
22–23 May 2017: Training
23–25 May 2017: Tutorials & Conference
London, UK

Reducing neural-network training time through hyperparameter optimization

Amitai Armon (Intel), Yahav Shadmi (Intel)
12:0012:30 Tuesday, 23 May 2017
Data science and advanced analytics
Location: London Suite 2/3
Secondary topics:  AI, Deep learning
Level: Intermediate
Average rating: *****
(5.00, 2 ratings)

Neural-network models have a set of configuration hyperparameters tuned to optimize a given model’s accuracy. While accurate models are the primary goal in machine learning, it is often necessary to reduce training time to avoid outdated models and maintain computational feasibility.

Yahav Shadmi demonstrates how to select hyperparameters to significantly reduce training time while maintaining accuracy, present examples for popular neural network models used for text and images, and describe a real-world optimization method for tuning.

This presentation is coauthored by Amitai Armon.

Photo of Amitai Armon

Amitai Armon


Amitai Armon is the chief data scientist for Intel’s Advanced Analytics group, which provides solutions for the company’s challenges in diverse domains ranging from design and manufacturing to sales and marketing, using machine learning and big data techniques. Previously, Amitai was the cofounder and director of research at TaKaDu, a provider of water-network analytics software to detect hidden underground leaks and network inefficiencies. The company received several international awards, including the World Economic Forum Technology Pioneers award. Amitai has about 15 years of experience in performing and leading data science work. He holds a PhD in computer science from the Tel Aviv University in Israel, where he previously completed his BSc (cum laude, at the age of 18).

Yahav Shadmi


Yahav Shadmi is a senior data scientist in Intel’s Advanced Analytics department, a group that provides solutions for diverse company challenges using machine learning and big data techniques, where he leads data science projects and solves data-driven problems in the CPU design and architecture domain. Yahav’s current research is on optimization acceleration of deep learning tasks. He holds an MSc in computer science and machine learning from the Haifa University, Israel.