Solving a problem with Deep Learning in a particular area depends on the availability of data. Large data sets are needed to make sure that the machine delivers desired results. So, is the big challenge consist to find a compromise between large data (quantity /quality) , overfitting and hyperparameters?
Dear Karim Zarour ,
As you have mentioned, a big problem in deep learning is the high demand for data (a requirement to fit the network capability). This issue prohibits us to use these approaches on small datasets. However, there exist more challenges in DL such as:
Well you are right artur. Another challenge is time complexity. And adoptive learning in real time is difficult.
Data to learn the past, model to extract pattern in data and hardware to run that model in a timely manner
- Badges
- Science topic
- Similar topics
- Automotive Engineering
- Machines