receiving sensor data from a sensor of the mobile device, wherein the sensor is operable to detect a plurality of different activities of a user of the mobile device;

from the sensor, collecting user-specific training data for a specific user of the mobile device performing the plurality of different activities;

generating a weight for each of the different activities based on the user-specific training data;

for the specific user, inputting the sensor data into a general model of the mobile device that outputs a likelihood result for each of the plurality of different activities, wherein the general model was trained based on sensor data from a plurality of different users performing the plurality of different activities;

for the specific user, applying each weight for each of its corresponding one of the different activities to adjust its corresponding likelihood result without modifying the general model and without retraining the general model;

determining a first one of the different activities is being performed by the specific user by selecting a highest value of the adjusted likelihood results; and

performing an action on the device for the specific user based on a determination of the first activity, wherein the different activities comprise walking, running, biking, and resting activities, wherein the action comprises turning on step detection and/or location detection for the mobile device if the first activity comprises walking, running or biking, wherein the action comprises turning off the step detection and/or location detection if the first activity is a resting activity.