Abstract | ||
---|---|---|
This work is motivated by the engineering task of achieving a near state-of-the-art face recognition on a minimal computing budget running on an embedded system. Our main technical contribution centers around a novel training method, called Multibatch, for similarity learning, i.e., for the task of generating an invariant ``face signatureu0027u0027 through training pairs of ``sameu0027u0027 and ``not-sameu0027u0027 face images. The Multibatch method first generates signatures for a mini-batch of $k$ face images and then constructs an unbiased estimate of the full gradient by relying on all $k^2-k$ pairs from the mini-batch. We prove that the variance of the Multibatch estimator is bounded by $O(1/k^2)$, under some mild conditions. In contrast, the standard gradient estimator that relies on random $k/2$ pairs has a variance of order $1/k$. The smaller variance of the Multibatch estimator significantly speeds up the convergence rate of stochastic gradient descent. Using the Multibatch method we train a deep convolutional neural network that achieves an accuracy of $98.2%$ on the LFW benchmark, while its prediction runtime takes only $30$msec on a single ARM Cortex A9 core. Furthermore, the entire training process took only 12 hours on a single Titan X GPU. |
Year | Venue | DocType |
---|---|---|
2016 | NIPS | Conference |
Volume | Citations | PageRank |
abs/1605.07270 | 9 | 0.50 |
References | Authors | |
1 | 5 |
Name | Order | Citations | PageRank |
---|---|---|---|
Oren Tadmor | 1 | 9 | 0.50 |
Yonatan Wexler | 2 | 805 | 47.62 |
Tal Rosenwein | 3 | 9 | 0.50 |
Shai Shalev-Shwartz | 4 | 3681 | 276.32 |
Amnon Shashua | 5 | 3396 | 384.93 |