In this post, I’ll just present in brief the main difference between Deep Learning and Machine Learning, besides, I’ll try to give you some concepts closely related to DL. Let’s start with:
Deep Learning is a subtype of Machine Learning focus on Images as data.
Transfer Learning
This type is based on the term Experience based Learning, in other words, the learning process is performed taking a previous learned parameters from some model trained in a dataset and then we train the same model with another dataset that has different distribution of classes or even other classes.
In summary, Transfer learning is when a model developed for one task is reused to work on a second task. Is called Domain Adaptation as well.
Fine Tunning
Is an approach of Transfer Learning, for example, we have a dataset that we want to train, so we can use 90% of it in training. Then we re-train the same model with the remaining 10%. Usually, we change the learning rate to a smaller one, so it does not have a significant impact on the already adjusted weights.
Deep Learning Related models (Nets)
Deep learning can be in supervised or unsupervised manner. Supervised learning and unsupervised are completely opposite of each other. In supervised learning task of inferring from labeled data and in unsupervised learning task of inferring from unlabeled data. Supervised learning includes classification and unsupervised learning includes pattern analysis.
Unsupervised Pretrained Networks
This architecture is based on Discriminative networks to solve the problem of domain or distribution learning, this group has three specific architectures:
- Autoencoders
- Deep Belief Networks (DBNs)
- Generative Adversarial Networks (GANs)
Recursive Neural Networks (RvNN)
This architecture is based on a shared-weight matrix and a binary tree structure that allows network to learn varying sequences of words or parts of an image. This architecture amis to produce an structured prediction.
Recurrent Neural Networks (RNN)
This architecture is based on ANN to solve more complex problems like sequential data (e.g. time series) implementing concepts like “forgot cells” and saving states to improve the learning process between each layer.
Convolutional Neural Networks (CNN)
This architecture is based on ANN to solve complex problems more likely data comming from images, this method aims to learn filters obtained by the convolution layers and relate features extracted with some label associated.
- Automatic speech recognition
- Image recognition
- Visual Art Processing
- Natural language processing
- Drug discovery and toxicology
- Customer relationship management
- Recommendation systems
- Bioinformatics
- Mobile Advertising