(a) 𝚍 𝚒 𝚖
(b) 𝚍 𝚒 𝚖
(c) 𝚍 𝚒 𝚖
(d) 𝚑 𝚊 𝚛 𝚙 𝚞 𝚛
(e) 𝚗 𝚖 𝚏 𝚍 𝚒 𝚟
Figure 8: (a)–(e) Example basis vectors learnt by each algorithm when trained on the CBCL Face Database, with 𝚍 𝚒 𝚖 . (f) The change during the course of training of the mean Euclidean distance between the input and the reconstructed input. Results show mean of 10 trials performed using each algorithm. The best and worst performance over these 10 trials is shown by the error bars (which are very small in each case). Note that the total training time used for each algorithm (and hence the meaning of the value “20" on the x-axis of this graph) was 200 epochs for 𝚗 𝚖 𝚏 𝚍 𝚒 𝚟 , 2000 epochs for algorithms 𝚗 𝚖 𝚏 𝚜 𝚌 and 𝚍 𝚒 𝚖 , and 20 epochs for all the other algorithms. Training times are therefore not all directly comparable: at any particular time 𝚏 𝚢 𝚏 𝚎 has seen 10 times more data and been updated 10 times more than 𝚗 𝚖 𝚏 𝚍 𝚒 𝚟 , while 𝚗 𝚖 𝚏 𝚜 𝚌 and 𝚗 𝚖 𝚏 𝚜 𝚎 𝚚 have seen 100 times more data but been updated 24 times less than 𝚍 𝚒 𝚖 .