This is very sound advice. But I'd like to encourage applied data scientists to learn about deep learning. It's a very powerful toolkit, very appropriate for many bio problems with large datasets requiring integration of diverse structured input data types. 1/
And the only way to learn to use NNs is by working with them. Do an actual project with them. Consult with someone familiar with NN to make sure that the problem/dataset is a good match for NNs. 2/
The frameworks built for NNs like Keras, Tensorflow, Pytorch, Edward are incredible and have very valuable functionality beyond just training NNs. 3/
Also remember that linear/logistic regressor (the artificial neuron) is the basic building block of a NN. The final layer of a classic deep NN performs linear/logistic regression but on a transformed version (that's what all those layers do) of your original input features. 4/
If your original features of your input data are expressive enough for your prediction problem, there is no need for complex transformations. A 0 layer NN (linear/logistic regression) will do just fine. 5/
If u have more complex structure in your input data (eg. an image, a DNA/RNA/protein sequence, a DNAse-seq profile) that you want to learn from, a NN will assist you in extracting predictive features from the raw input before feeding it to linear/logistic regression. 6/
Also worth noting that 'deep' NNs, do not need to have 200 layers of neurons. U r supposed to & shud perform an 'architecture search' to identify the optimal complexity of ur model. The simplest NN model is a linear/logistic regressor. If that works great, y add more layers? 7/
There is a lot of hype and anti-hype about deep learning. Don't blindly fall for the propaganda from either side. Stay in the middle. Learn it as another toolkit in your arsenal. Learn it's strengths and weaknesses. Use it where appropriate. 8/8
