We had a phenomenal week at Deep Learning World 2019 in Vegas.
Deep Learning World, co-located alongside four Predictive Analytics World events, is the premier conference covering the commercial deployment of deep learning. With more than 800 attendees and 150+ sessions all focused on deep learning, predictive analytics, and ways these technologies can improve business as we know it, it's the place to see and be seen in the ML/DL circles.
The conference was quite interesting with many topics that are relevant in the predictive analytics world. One thing that resonated across all talks and conversations: everyone is looking for ways to enhance the customer experience, and machine learning is driving that improvement.
Some of the interesting highlights include:
- A common trend, expressed also in a great keynote by Mohammad Shokoohi-Yekta, is that Time Series is the new Big Data, and it currently needs Data Science teams to act as the bridge between data and providing the value for the organization.
- PayPal presented how they are using machine learning and alerts in attempt to reach 5 9s availability for their network.
- Uber is using machine learning to improve their app testing methodologies, recognizing the opportunity to model against actual app performance and environmental data from the phone.
- Isuzu is using deep learning to predict failures before they happen, giving enough time to address them and explaining the root cause to the technicians.
Deep Learning and Automated Pattern Discovery
As for our own co-founder & VP of R&D Przemek Maciolek's presentation, it was one of the most well attended talks in the Deep Learning track, and the feedback was both outstanding and encouraging.
A discussion followed, where some interesting questions were asked. One of the questions was asking for the use of word embeddings, which is an interesting technique used in Natural Language Processing (NLP). Interesting enough but in the case of logs, and particularly with the LogSense automated patterned discovery, this technique is rendered unnecessary since LogSense automatically converts unstructured data into structured data. Another question asked about managing the patterns space. This is an interesting question, and one that has a fairly simple answer: each LogSense customer gets its own space (which is also true for all other data), which keeps the patterns separate and focused for each individual customer environment. We'll share more details in the coming weeks and months as we continue to follow-up on these conversations.
For now, thanks to everyone who came out for Przemek's talk, and for all the great conversations during the week. We're thrilled for what the future holds for ML/DL and particularly the role LogSense will play in that success.
If you're interested in taking a closer look at the LogSense platform, sign up for your free trial today!