Logistic Regression looks at outputting a binary choice (1 or 0, Yes or No, etc) by determining a probability or "confidence". The technique is popularly used for classification based problems (ex: Is the tumor malignant?) where the output choice is given a probability of being correct (ex: 85% likely that the answer is "yes" so output "yes"). Linear Regression has a virtually unlimited number of answers/outputs since it usually outputs a forecasted value (ex: What value will this building have in 2019?). Linear Regression outputs a likely value in a distribution of potential values. The outputs are continuous as opposed to discrete and it is often used for problems like forecasting.