tinyML Talks: Build an Edge optimized tinyML application for the Arduino Nano 33 BLE Sense How Chris Knorowski, CTO of Sensiml goes about tinyML as a solution for a problem 1. Is the solution complex enough to warranty ML? 2. Does the data need to be on the device vs cloud? security reasons. 3. Business case Whats your sample rate and how will you collect the data? Wi-Fi or BLE? While adding a sensor for every golf club might not make sense, there is a company which has some intelligence in golf ball with GPS and Graf ball with analytics ability. 4. Data Collection - Variance of data to decide how much data to collect, domain expertise for labelling data. 12:43 Good question on the order of fitting the code vs shining up the model Important to have negative examples when building gesture applications. Understanding Confusion matrix with the doctor, patient and pregnancy. This meme explains is better, the confusion matrix shows us...