Skip to main content

Table 2 Fusion functions for learning drug–target edge representations

From: DTi2Vec: Drug–target interaction prediction using network embedding and ensemble learning

Fusion function

Function equation

1. Concatenation

\(\left[ { f\left( d \right) + f\left( t \right) } \right] = f\left( d \right) f\left( t \right)\)

2. Hadamard

[f(d) f(t)]\(= f\left( d \right) * f\left( t \right)\)

3. Average

[f(d) f(t)] \(= \left( {f\left( d \right) + f\left( t \right)} \right)/2\)

4. Weighted-L1

\(\left\| { f\left( d \right) . f\left( t \right)} \right\|_{1} = \left| {f\left( d \right) - f\left( t \right)} \right|\)

5. Weighted-L2

\(\left\| {f\left( d \right) . f\left( t \right)} \right\|_{2} = \left| {f\left( d \right) - f\left( t \right)} \right|^{2}\)

  1. represents element wise matrix multiplication, and represents the average of the two matrices