CS计算机代考程序代写 database Bayesian data mining deep learning algorithm Published as a conference paper at ICLR 2017

Published as a conference paper at ICLR 2017
ON LARGE-BATCH TRAINING FOR DEEP LEARNING: GENERALIZATION GAP AND SHARP MINIMA
Nitish Shirish Keskar∗
Northwestern University
Evanston, IL 60208 keskar.nitish@u.northwestern.edu
Jorge Nocedal
Northwestern University
Evanston, IL 60208 j-nocedal@northwestern.edu
Ping Tak Peter Tang
Intel Corporation
Santa Clara, CA 95054 peter.tang@intel.com
Dheevatsa Mudigere
Intel Corporation
Bangalore, India dheevatsa.mudigere@intel.com
Mikhail Smelyanskiy
Intel Corporation
Santa Clara, CA 95054 mikhail.smelyanskiy@intel.com
ABSTRACT
The stochastic gradient descent (SGD) method and its variants are algorithms of choice for many Deep Learning tasks. These methods operate in a small-batch regime wherein a fraction of the training data, say 32–512 data points, is sampled to compute an approximation to the gradient. It has been observed in practice that when using a larger batch there is a degradation in the quality of the model, as measured by its ability to generalize. We investigate the cause for this generaliza- tion drop in the large-batch regime and present numerical evidence that supports the view that large-batch methods tend to converge to sharp minimizers of the training and testing functions—and as is well known, sharp minima lead to poorer generalization. In contrast, small-batch methods consistently converge to flat min- imizers, and our experiments support a commonly held view that this is due to the inherent noise in the gradient estimation. We discuss several strategies to attempt to help large-batch methods eliminate this generalization gap.
1 INTRODUCTION
Deep Learning has emerged as one of the cornerstones of large-scale machine learning. Deep Learn- ing models are used for achieving state-of-the-art results on a wide variety of tasks including com- puter vision, natural language processing and reinforcement learning; see (Bengio et al., 2016) and the references therein. The problem of training these networks is one of non-convex optimization. Mathematically, this can be represented as:
1 􏰈M
min f(x) := fi(x), (1)
x∈Rn M
i=1
where fi is a loss function for data point i ∈ {1, 2, · · · , M } which captures the deviation of the model prediction from the data, and x is the vector of weights being optimized. The process of optimizing this function is also called training of the network. Stochastic Gradient Descent (SGD) (Bottou, 1998; Sutskever et al., 2013) and its variants are often used for training deep networks.
∗Work was performed when author was an intern at Intel Corporation 1
arXiv:1609.04836v2 [cs.LG] 9 Feb 2017

Published as a conference paper at ICLR 2017
These methods minimize the objective function f by iteratively taking steps of the form: 􏰅1􏰆
x =x−α 􏰈∇f(x) , (2) k+1 k k |Bk| i k
i∈Bk
whereBk ⊂{1,2,···,M}isthebatchsampledfromthedatasetandαk isthestepsizeatiteration k. These methods can be interpreted as gradient descent using noisy gradients, which and are often referred to as mini-batch gradients with batch size |Bk|. SGD and its variants are employed in a small-batch regime, where |Bk | ≪ M and typically |Bk | ∈ {32, 64, · · · , 512}. These configura- tions have been successfully used in practice for a large number of applications; see e.g. (Simonyan & Zisserman, 2014; Graves et al., 2013; Mnih et al., 2013). Many theoretical properties of these methods are known. These include guarantees of: (a) convergence to minimizers of strongly-convex functions and to stationary points for non-convex functions (Bottou et al., 2016), (b) saddle-point avoidance (Ge et al., 2015; Lee et al., 2016), and (c) robustness to input data (Hardt et al., 2015).
Stochastic gradient methods have, however, a major drawback: owing to the sequential nature of the iteration and small batch sizes, there is limited avenue for parallelization. While some efforts have been made to parallelize SGD for Deep Learning (Dean et al., 2012; Das et al., 2016; Zhang et al., 2015), the speed-ups and scalability obtained are often limited by the small batch sizes. One natu- ral avenue for improving parallelism is to increase the batch size |Bk|. This increases the amount of computation per iteration, which can be effectively distributed. However, practitioners have ob- served that this leads to a loss in generalization performance; see e.g. (LeCun et al., 2012). In other words, the performance of the model on testing data sets is often worse when trained with large- batch methods as compared to small-batch methods. In our experiments, we have found the drop in generalization (also called generalization gap) to be as high as 5% even for smaller networks.
In this paper, we present numerical results that shed light into this drawback of large-batch methods. We observe that the generalization gap is correlated with a marked sharpness of the minimizers obtained by large-batch methods. This motivates efforts at remedying the generalization problem, as a training algorithm that employs large batches without sacrificing generalization performance would have the ability to scale to a much larger number of nodes than is possible today. This could potentially reduce the training time by orders-of-magnitude; we present an idealized performance model in the Appendix C to support this claim.
The paper is organized as follows. In the remainder of this section, we define the notation used in this paper, and in Section 2 we present our main findings and their supporting numerical evidence. In Section 3 we explore the performance of small-batch methods, and in Section 4 we briefly discuss the relationship between our results and recent theoretical work. We conclude with open questions concerning the generalization gap, sharp minima, and possible modifications to make large-batch training viable. In Appendix E, we present some attempts to overcome the problems of large-batch training.
1.1 NOTATION
We use the notation fi to denote the composition of loss function and a prediction function corre- sponding to the ith data point. The vector of weights is denoted by x and is subscripted by k to denote an iteration. We use the term small-batch (SB) method to denote SGD, or one of its variants like ADAM (Kingma & Ba, 2015) and ADAGRAD (Duchi et al., 2011), with the proviso that the gradient approximation is based on a small mini-batch. In our setup, the batch Bk is randomly sam- pled and its size is kept fixed for every iteration. We use the term large-batch (LB) method to denote any training algorithm that uses a large mini-batch. In our experiments, ADAM is used to explore the behavior of both a small or a large batch method.
2 DRAWBACKS OF LARGE-BATCH METHODS
2.1 OUR MAIN OBSERVATION
As mentioned in Section 1, practitioners have observed a generalization gap when using large-batch methods for training deep learning models. Interestingly, this is despite the fact that large-batch methods usually yield a similar value of the training function as small-batch methods. One may put
2

Published as a conference paper at ICLR 2017
forth the following as possible causes for this phenomenon: (i) LB methods over-fit the model; (ii) LB methods are attracted to saddle points; (iii) LB methods lack the explorative properties of SB methods and tend to zoom-in on the minimizer closest to the initial point; (iv) SB and LB methods converge to qualitatively different minimizers with differing generalization properties. The data presented in this paper supports the last two conjectures.
The main observation of this paper is as follows:
The concept of sharp and flat minimizers have been discussed in the statistics and machine learning literature. (Hochreiter & Schmidhuber, 1997) (informally) define a flat minimizer x ̄ as one for which the function varies slowly in a relatively large neighborhood of x ̄. In contrast, a sharp minimizer xˆ is such that the function increases rapidly in a small neighborhood of xˆ. A flat minimum can be de- scribed with low precision, whereas a sharp minimum requires high precision. The large sensitivity of the training function at a sharp minimizer negatively impacts the ability of the trained model to generalize on new data; see Figure 1 for a hypothetical illustration. This can be explained through the lens of the minimum description length (MDL) theory, which states that statistical models that require fewer bits to describe (i.e., are of low complexity) generalize better (Rissanen, 1983). Since flat minimizers can be specified with lower precision than to sharp minimizers, they tend to have bet- ter generalization performance. Alternative explanations are proffered through the Bayesian view of learning (MacKay, 1992), and through the lens of free Gibbs energy; see e.g. Chaudhari et al. (2016).
The lack of generalization ability is due to the fact that large-batch methods tend to converge to sharp minimizers of the training function. These minimizers are characterized by a signif- icant number of large positive eigenvalues in ∇2f(x), and tend to generalize less well. In contrast, small-batch methods converge to flat minimizers characterized by having numerous small eigenvalues of ∇2f(x). We have observed that the loss function landscape of deep neural networks is such that large-batch methods are attracted to regions with sharp minimizers and that, unlike small-batch methods, are unable to escape basins of attraction of these minimizers.
Training Function
f (x)
Flat Minimum Sharp Minimum
Figure 1: A Conceptual Sketch of Flat and Sharp Minima. The Y-axis indicates value of the loss function and the X-axis the variables (parameters)
2.2 NUMERICAL EXPERIMENTS
In this section, we present numerical results to support the observations made above. To this end, we make use of the visualization technique employed by (Goodfellow et al., 2014b) and a proposed heuristic metric of sharpness (Equation (4)). We consider 6 multi-class classification network con- figurations for our experiments; they are described in Table 1. The details about the data sets and network configurations are presented in Appendices A and B respectively. As is common for such problems, we use the mean cross entropy loss as the objective function f .
The networks were chosen to exemplify popular configurations used in practice like AlexNet (Krizhevsky et al., 2012) and VGGNet (Simonyan & Zisserman, 2014). Results on other networks
3
Testing Function

Published as a conference paper at ICLR 2017
Table
Name Network Type
F1 Fully Connected
F2 Fully Connected
C1 (Shallow) Convolutional
C2 (Deep) Convolutional
C3 (Shallow) Convolutional
C4 (Deep) Convolutional
1: Network Configurations
Architecture Section B.1 Section B.2 Section B.3 Section B.4 Section B.3 Section B.4
Data set
MNIST (LeCun et al., 1998a)
TIMIT (Garofolo et al., 1993)
CIFAR-10 (Krizhevsky & Hinton, 2009) CIFAR-10
CIFAR-100 (Krizhevsky & Hinton, 2009) CIFAR-100
and using other initialization strategies, activation functions, and data sets showed similar behavior. Since the goal of our work is not to achieve state-of-the-art accuracy or time-to-solution on these tasks but rather to characterize the nature of the minima for LB and SB methods, we only describe the final testing accuracy in the main paper and ignore convergence trends.
For all experiments, we used 10% of the training data as batch size for the large-batch experiments and 256 data points for small-batch experiments. We used the ADAM optimizer for both regimes. Experiments with other optimizers for the large-batch experiments, including ADAGRAD (Duchi et al., 2011), SGD (Sutskever et al., 2013) and adaQN (Keskar & Berahas, 2016), led to similar results. All experiments were conducted 5 times from different (uniformly distributed random) starting points and we report both mean and standard-deviation of measured quantities. The baseline performance for our setup is presented Table 2. From this, we can observe that on all networks, both approaches led to high training accuracy but there is a significant difference in the generalization performance. The networks were trained, without any budget or limits, until the loss function ceased to improve.
Table 2: Performance of small-batch (SB) and large-batch (LB) variants of ADAM on the 6 networks listed in Table 1
Training Accuracy Name SB LB
Testing Accuracy SB LB
F1 99.66% ± 0.05%
F2 99.99% ± 0.03%
C1 99.89% ± 0.02%
C2 99.99% ± 0.04%
C3 99.56% ± 0.44%
C4 99.10% ± 1.23%
99.92% 98.35% 99.66% 99.99% 99.88% 99.57%
± 0.01% ± 2.08% ± 0.2% ± 0.01% ± 0.30% ± 1.84%
98.03% ± 0.07% 64.02% ± 0.2% 80.04% ± 0.12% 89.24% ± 0.12% 49.58% ± 0.39% 63.08% ± 0.5%
97.81% ± 0.07% 59.45% ± 1.05% 77.26% ± 0.42% 87.26% ± 0.07% 46.45% ± 0.43% 57.81% ± 0.17%
We emphasize that the generalization gap is not due to over-fitting or over-training as commonly observed in statistics. This phenomenon manifest themselves in the form of a testing accuracy curve that, at a certain iterate peaks, and then decays due to the model learning idiosyncrasies of the training data. This is not what we observe in our experiments; see Figure 2 for the training–testing curve of the F2 and C1 networks, which are representative of the rest. As such, early-stopping heuristics aimed at preventing models from over-fitting would not help reduce the generalization gap. The difference between the training and testing accuracies for the networks is due to the specific choice of the network (e.g. AlexNet, VGGNet etc.) and is not the focus of this study. Rather, our goal is to study the source of the testing performance disparity of the two regimes, SB and LB, on a given network model.
2.2.1 PARAMETRIC PLOTS
We first present parametric 1-D plots of the function as described in (Goodfellow et al., 2014b). Let x⋆s and x⋆l indicate the solutions obtained by running ADAM using small and large batch sizes respectively. We plot the loss function, on both training and testing data sets, along a line-segment containing the two points. Specifically, for α ∈ [−1, 2], we plot the function f (αx⋆l + (1 − α)x⋆s ) and also superimpose the classification accuracy at the intermediate points; see Figure 31. For this
1The code to reproduce the parametric plot on exemplary networks can be found in our GitHub repository: https://github.com/keskarnitish/large-batch-training.
4

Published as a conference paper at ICLR 2017
100 90 80 70 60 50 40 30 20 10
0 20 40 60 80 100 Epoch
SB – Training SB – Testing LB – Training LB – Testing
100 90 80 70 60 50 40 30 20
0 20 40 60 80 100 Epoch
SB – Training SB – Testing LB – Training LB – Testing
(a) Network F2 (b) Network C1
Figure 2: Training and testing accuracy for SB and LB methods as a function of epochs.
experiment, we randomly chose a pair of SB and LB minimizers from the 5 trials used to generate
the data in Table 2. The plots show that the LB minima are strikingly sharper than the SB minima
in this one-dimensional manifold. The plots in Figure 3 only explore a linear slice of the function,
but in Figure 7 in Appendix D, we plot f (sin( απ )x⋆ + cos( απ )x⋆ ) to monitor the function along a 2l 2s
curved path between the two minimizers . There too, the relative sharpness of the minima is evident. 2.2.2 SHARPNESS OF MINIMA
So far, we have used the term sharp minimizer loosely, but we noted that this concept has received attention in the literature (Hochreiter & Schmidhuber, 1997). Sharpness of a minimizer can be characterized by the magnitude of the eigenvalues of ∇2f(x), but given the prohibitive cost of this computation in deep learning applications, we employ a sensitivity measure that, although imperfect, is computationally feasible, even for large networks. It is based on exploring a small neighborhood of a solution and computing the largest value that the function f can attain in that neighborhood. We use that value to measure the sensitivity of the training function at the given local minimizer. Now, since the maximization process is not accurate, and to avoid being mislead by the case when a large value of f is attained only in a tiny subspace of Rn, we perform the maximization both in the entire space Rn as well as in random manifolds. For that purpose, we introduce an n × p matrix A, whose columns are randomly generated. Here p determines the dimension of the manifold, which in our experiments is chosen as p = 100.
Specifically, let Cε denote a box around the solution over which the maximization of f is performed, and let A ∈ Rn×p be the matrix defined above. In order to ensure invariance of sharpness to problem dimension and sparsity, we define the constraint set Cε as:
Cε ={z∈Rp :−ε(|(A+x)i|+1)≤zi ≤ε(|(A+x)i|+1) ∀i∈{1,2,···,p}}, (3) where A+ denotes the pseudo-inverse of A. Thus ε controls the size of the box. We can now define
our measure of sharpness (or sensitivity).
Metric 2.1. Given x ∈ Rn, ε > 0 and A ∈ Rn×p, we define the (Cε, A)-sharpness of f at x as: φx,f(ε,A):=(maxy∈Cε f(x+Ay))−f(x)×100. (4)
1+f(x)
Unless specified otherwise, we use this metric for sharpness for the rest of the paper; if A is not spec- ified, it is assumed to be the identity matrix, In. (We note in passing that, in the convex optimization literature, the term sharp minimum has a different definition (Ferris, 1988), but that concept is not useful for our purposes.)
In Tables 3 and 4, we present the values of the sharpness metric (4) for the minimizers of the various problems. Table 3 explores the full-space (i.e., A = In) whereas Table 4 uses a randomly sampled n × 100 dimensional matrix A. We report results with two values of ε, (10−3 , 5 · 10−4 ). In all experiments, we solve the maximization problem in Equation (4) inexactly by applying 10 iterations of L-BFGS-B (Byrd et al., 1995). This limit on the number of iterations was necessitated by the
5
Accuracy
Accuracy

Published as a conference paper at ICLR 2017
12 100
Train Test
10 8 6 4 2
80 60 40 20
00 1.0 0.5 0.0 0.5 1.0 1.5 2.0
alpha
14 100 12
10
8
6
4
2
Train Test
80 60 40 20
00 1.0 0.5 0.0 0.5 1.0 1.5 2.0
alpha
(a) F1
(b) F2
8 7 6 5 4 3 2 1 0
1.0
0.5
0.0
0.5 alpha
1.0
100 90 80 70 60 50 40 30 20 10
1.5 2.0
Train Test
7 100 6
5
4
3
2
1
00 1.0 0.5 0.0 0.5 1.0 1.5 2.0
alpha
Train Test
80 60 40 20
(c) C1
(d) C2
10 100
8 80
6 60
4 40
2 20
00 1.0 0.5 0.0 0.5 1.0 1.5 2.0
alpha
Train Test
12 100
Train Test
10 8 6 4 2
80 60 40 20
00 1.0 0.5 0.0 0.5 1.0 1.5 2.0
alpha
(e) C3
(f) C4
Figure 3: Parametric Plots – Linear (Left vertical axis corresponds to cross-entropy loss, f, and right vertical axis corresponds to classification accuracy; solid line indicates training data set and dashed line indicated testing data set); α = 0 corresponds to the SB minimizer and α = 1 to the LB minimizer.
large cost of evaluating the true objective f. Both tables show a 1–2 order-of-magnitude difference between the values of our metric for the SB and LB regimes. These results reinforce the view that the solutions obtained by a large-batch method defines points of larger sensitivity of the training function. In Appedix E, we describe approaches to attempt to remedy this generalization problem of LB methods. These approaches include data augmentation, conservative training and adversarial training. Our preliminary findings show that these approaches help reduce the generalization gap but still lead to relatively sharp minimizers and as such, do not completely remedy the problem.
Note that Metric 2.1 is closely related to the spectrum of ∇2f(x). Assuming ε to be small enough, when A = In, the value (4) relates to the largest eigenvalue of ∇2f(x) and when A is randomly sampled it approximates the Ritz value of ∇2f(x) projected onto the column-space of A.
6
Cross Entropy
Cross Entropy
Cross Entropy
Accuracy
Accuracy
Accuracy
Cross Entropy
Cross Entropy
Cross Entropy
Accuracy
Accuracy
Accuracy

Published as a conference paper at ICLR 2017
Table 3: Sharpness of Minima in Full Space; ε is defined in (3).
ε = 10−3 ε = 5·10−4 SB LB SB LB
F1 F2 C1 C2 C3 C4
1.23 ± 0.83
1.39 ± 0.02 28.58 ± 3.13 8.68 ± 1.32 29.85 ± 5.98 12.83 ± 3.84
205.14 ± 69.52 310.64 ± 38.46 707.23 ± 43.04 925.32 ± 38.29 258.75 ± 8.96 421.84 ± 36.97
0.61 ± 0.27 0.90 ± 0.05 7.08 ± 0.88 2.07 ± 0.86 8.56 ± 0.99 4.07 ± 0.87
42.90 ± 17.14 93.15 ± 6.81 227.31 ± 23.23 175.31 ± 18.28 105.11 ± 13.22 109.35 ± 16.57
Table 4: Sharpness of Minima in Random Subspaces of Dimension 100
ε = 10−3 ε = 5·10−4 SB LB SB LB
F1 F2 C1 C2 C3 C4
0.11 ± 0.00 0.29 ± 0.02 2.18 ± 0.23 0.95 ± 0.34 17.02 ± 2.20 6.05 ± 1.13
9.22 ± 0.56 23.63 ± 0.54 137.25 ± 21.60 25.09 ± 2.61 236.03 ± 31.26 72.99 ± 10.96
0.05 ± 0.00 0.05 ± 0.00 0.71 ± 0.15 0.31 ± 0.08 4.03 ± 1.45 1.89 ± 0.33
9.17 ± 0.14
6.28 ± 0.19 29.50 ± 7.48 5.82 ± 0.52 86.96 ± 27.39 19.85 ± 4.12
We conclude this
resemble a cone, i.e., the function does not increase rapidly along all (or even most) directions. By sampling the loss function in a neighborhood of LB solutions, we observe that it rises steeply only along a small dimensional subspace (e.g. 5% of the whole space); on most other directions, the function is relatively flat.
3 SUCCESS OF SMALL-BATCH METHODS
It is often reported that when increasing the batch size for a problem, there exists a threshold after which there is a deterioration in the quality of the model. This behavior can be observed for the F2 and C1 networks in Figure 4. In both of these experiments, there is a batch size (≈ 15000 for F2 and ≈ 500 for C1) after which there is a large drop in testing accuracy. Notice also that the upward drift in value of the sharpness is considerably reduced around this threshold. Similar thresholds exist for the other networks in Table 1.
Let us now consider the behavior of SB methods, which use noisy gradients in the step computation. From the results reported in the previous section, it appears that noise in the gradient pushes the iterates out of the basin of attraction of sharp minimizers and encourages movement towards a flatter minimizer where noise will not cause exit from that basin. When the batch size is greater than the threshold mentioned above, the noise in the stochastic gradient is not sufficient to cause ejection from the initial basin leading to convergence to sharper a minimizer.
To explore that in more detail, consider the following experiment. We train the network for 100 epochs using ADAM with a batch size of 256, and retain the iterate after each epoch in memory. Using these 100 iterates as starting points we train the network using a LB method for 100 epochs and receive a 100 piggybacked (or warm-started) large-batch solutions. We plot in Figure 5 the testing accuracy and sharpness of these large-batch solutions, along with the testing accuracy of the small-batch iterates. Note that when warm-started with only a few initial epochs, the LB method does not yield a generalization improvement. The concomitant sharpness of the iterates also stays high. On the other hand, after certain number of epochs of warm-starting, the accuracy improves and sharpness of the large-batch iterates drop. This happens, apparently, when the SB method has ended its exploration phase and discovered a flat minimizer; the LB method is then able to converge towards it, leading to good testing accuracy.
It has been speculated that LB methods tend to be attracted to minimizers close to the starting point x0, whereas SB methods move away and locate minimizers that are farther away. Our numerical
section by noting that the sharp minimizers identified in our experiments do not
7

Published as a conference paper at ICLR 2017
65 64 63 62 61 60 59 58
0
10000
20000
30000
Batch Size
50000
1400 1200 1000 800 600 400 200 0
60000 70000
1E-
5E-4
3
40000
79.5 79.0 78.5 78.0 77.5 77.0 76.5 76.0 75.5
0
1000
2000
Batch Size
800 700 600 500 400 300 200 100 0
4000 5000
1E-3 5E-4
3000
(a) F2 (b) C1
Figure 4: Testing Accuracy and Sharpness v/s Batch Size. The X-axis corresponds to the batch size used for training the network for 100 epochs, left Y-axis corresponds to the testing accuracy at the final iterate and right Y-axis corresponds to the sharpness of that iterate. We report sharpness for two values of ε: 10−3 and 5 · 10−4.
SB
Piggyback LB
70 60 50 40 30
0 20 40 60 80 100 103
102 1E-3
101 100 10-1 10-2 10-3 10-4
5E-3
0 20 40 60 80 100 Epoch of SB Training
(a) F2
Figure 5: Warm-starting experiments. The upper figures report the testing accuracy of the SB method (blue line) and the testing accuracy of the warm started (piggybacked) LB method (red line), as a function of the number of epochs of the SB method. The lower figures plot the sharpness mea- sure (4) for the solutions obtained by the piggybacked LB method v/s the number of warm-starting epochs of the SB method.
8
85 80 75 70 65 60
SB Piggyback LB
0 20 40 60 80 100 103
102 101 100 10-1 10-2
1E-3 5E-3
0 20 40 60 80 100 Epoch of SB Training
(b) C1
Sharpness
Testing Accuracy
Testing Accuracy
Sharpness
Testing Accuracy
Testing Accuracy
Sharpness
Sharpness

Published as a conference paper at ICLR 2017
104 103 102 101 100
LB – 1E-3 LB – 5E-4 SB – 1E-3 SB – 5E-4
100
10-1
Cross Entropy
(a) F2
Figure 6: Sharpness v/s Cross Entropy Loss for SB and LB methods.
experiments support this view: we observed that the ratio of ∥x⋆s − x0∥2 and ∥x⋆l − x0∥2 was in the range of 3–10.
In order to further illustrate the qualitative difference between the solutions obtained by SB and LB methods, we plot in Figure 6 our sharpness measure (4) against the loss function (cross entropy) for one random trial of the F2 and C1 networks. For larger values of the loss function, i.e., near the initial point, SB and LB method yield similar values of sharpness. As the loss function reduces, the sharpness of the iterates corresponding to the LB method rapidly increases, whereas for the SB method the sharpness stays relatively constant initially and then reduces, suggesting an exploration phase followed by convergence to a flat minimizer.
4 DISCUSSION AND CONCLUSION
In this paper, we present numerical experiments that support the view that convergence to sharp minimizers gives rise to the poor generalization of large-batch methods for deep learning. To this end, we provide one-dimensional parametric plots and perturbation (sharpness) measures for a vari- ety of deep learning architectures. In Appendix E, we describe our attempts to remedy the problem, including data augmentation, conservative training and robust optimization. Our preliminary inves- tigation suggests that these strategies do not correct the problem; they improve the generalization of large-batch methods but still lead to relatively sharp minima. Another prospective remedy includes the use of dynamic sampling where the batch size is increased gradually as the iteration progresses (Byrd et al., 2012; Friedlander & Schmidt, 2012). The potential viability of this approach is sug- gested by our warm-starting experiments (see Figure 5) wherein high testing accuracy is achieved using a large-batch method that is warm-start with a small-batch method.
Recently, a number of researchers have described interesting theoretical properties of the loss sur- face of deep neural networks; see e.g. (Choromanska et al., 2015; Soudry & Carmon, 2016; Lee et al., 2016). Their work shows that, under certain regularity assumptions, the loss function of deep learning models is fraught with many local minimizers and that many of these minimizers corre- spond to a similar loss function value. Our results are in alignment these observations since, in our experiments, both sharp and flat minimizers have very similar loss function values. We do not know, however, if the theoretical models mentioned above provide information about the existence and density of sharp minimizers of the loss surface.
Our results suggest some questions: (a) can one prove that large-batch (LB) methods typically con- verge to sharp minimizers of deep learning training functions? (In this paper, we only provided some numerical evidence.); (b) what is the relative density of the two kinds of minima?; (c) can one design neural network architectures for various tasks that are suitable to the properties of LB methods?; (d) can the networks be initialized in a way that enables LB methods to succeed?; (e) is it possible, through algorithmic or regulatory means to steer LB methods away from sharp minimizers?
9
103
102
101
100
LB – 1E-3 LB – 5E-4 SB – 1E-3 SB – 5E-4
100
Cross Entropy
(b) C1
10-1
Sharpness
Sharpness

Published as a conference paper at ICLR 2017
REFERENCES
Yoshua Bengio, Ian Goodfellow, and Aaron Courville. Deep learning. Book in preparation for MIT
Press, 2016. URL http://www.deeplearningbook.org.
Dimitris Bertsimas, Omid Nohadani, and Kwong Meng Teo. Robust optimization for unconstrained
simulation-based problems. Operations Research, 58(1):161–178, 2010.
Le ́on Bottou. Online learning and stochastic approximations. On-line learning in neural networks,
17(9):142, 1998.
Le ́on Bottou, Frank E Curtis, and Jorge Nocedal. Optimization methods for large-scale machine
learning. arXiv preprint arXiv:1606.04838, 2016.
Richard H Byrd, Peihuang Lu, Jorge Nocedal, and Ciyou Zhu. A limited memory algorithm for
bound constrained optimization. SIAM Journal on Scientific Computing, 16(5):1190–1208, 1995. Richard H Byrd, Gillian M Chin, Jorge Nocedal, and Yuchen Wu. Sample size selection in opti-
mization methods for machine learning. Mathematical programming, 134(1):127–155, 2012. Pratik Chaudhari, Anna Choromanska, Stefano Soatto, and Yann LeCun. Entropy-sgd: Biasing
gradient descent into wide valleys. arXiv preprint arXiv:1611.01838, 2016.
Anna Choromanska, Mikael Henaff, Michael Mathieu, Ge ́rard Ben Arous, and Yann LeCun. The
loss surfaces of multilayer networks. In AISTATS, 2015.
Dipankar Das, Sasikanth Avancha, Dheevatsa Mudigere, Karthikeyan Vaidynathan, Srinivas Srid- haran, Dhiraj Kalamkar, Bharat Kaul, and Pradeep Dubey. Distributed deep learning using syn- chronous stochastic gradient descent. arXiv preprint arXiv:1602.06709, 2016.
Jeffrey Dean, Greg Corrado, Rajat Monga, Kai Chen, Matthieu Devin, Mark Mao, Andrew Senior, Paul Tucker, Ke Yang, Quoc V Le, et al. Large scale distributed deep networks. In Advances in neural information processing systems, pp. 1223–1231, 2012.
J. Duchi, E. Hazan, and Y. Singer. Adaptive subgradient methods for online learning and stochastic optimization. The Journal of Machine Learning Research, 12:2121–2159, 2011.
Michael Charles Ferris. Weak sharp minima and penalty functions in mathematical programming. PhD thesis, University of Cambridge, 1988.
Michael P Friedlander and Mark Schmidt. Hybrid deterministic-stochastic methods for data fitting. SIAM Journal on Scientific Computing, 34(3):A1380–A1405, 2012.
John S Garofolo, Lori F Lamel, William M Fisher, Jonathan G Fiscus, David S Pallett, Nancy L Dahlgren, and Victor Zue. Timit acoustic-phonetic continuous speech corpus. Linguistic data consortium, Philadelphia, 33, 1993.
Rong Ge, Furong Huang, Chi Jin, and Yang Yuan. Escaping from saddle pointsonline stochastic gradient for tensor decomposition. In Proceedings of The 28th Conference on Learning Theory, pp. 797–842, 2015.
Ian J Goodfellow, Jonathon Shlens, and Christian Szegedy. Explaining and harnessing adversarial examples. arXiv preprint arXiv:1412.6572, 2014a.
Ian J Goodfellow, Oriol Vinyals, and Andrew M Saxe. Qualitatively characterizing neural network optimization problems. arXiv preprint arXiv:1412.6544, 2014b.
Alex Graves, Abdel-rahman Mohamed, and Geoffrey Hinton. Speech recognition with deep recur- rent neural networks. In 2013 IEEE international conference on acoustics, speech and signal processing, pp. 6645–6649. IEEE, 2013.
M. Hardt, B. Recht, and Y. Singer. Train faster, generalize better: Stability of stochastic gradient descent. arXiv preprint arXiv:1509.01240, 2015.
Sepp Hochreiter and Ju ̈rgen Schmidhuber. Flat minima. Neural Computation, 9(1):1–42, 1997. 10

Published as a conference paper at ICLR 2017
Sergey Ioffe and Christian Szegedy. Batch normalization: Accelerating deep network training by reducing internal covariate shift. arXiv preprint arXiv:1502.03167, 2015.
Nitish Shirish Keskar and Albert S. Berahas. adaQN: An Adaptive Quasi-Newton Algorithm for Training RNNs, pp. 1–16. Springer International Publishing, Cham, 2016.
D. Kingma and J. Ba. Adam: A method for stochastic optimization. In International Conference on Learning Representations (ICLR 2015), 2015.
Alex Krizhevsky and Geoffrey Hinton. Learning multiple layers of features from tiny images. 2009.
Alex Krizhevsky, Ilya Sutskever, and Geoffrey E Hinton. Imagenet classification with deep convo- lutional neural networks. In Advances in neural information processing systems, pp. 1097–1105, 2012.
Yann LeCun, Le ́on Bottou, Yoshua Bengio, and Patrick Haffner. Gradient-based learning applied to document recognition. Proceedings of the IEEE, 86(11):2278–2324, 1998a.
Yann LeCun, Corinna Cortes, and Christopher JC Burges. The mnist database of handwritten digits, 1998b.
Yann A LeCun, Le ́on Bottou, Genevieve B Orr, and Klaus-Robert Mu ̈ller. Efficient backprop. In Neural networks: Tricks of the trade, pp. 9–48. Springer, 2012.
Jason D Lee, Max Simchowitz, Michael I Jordan, and Benjamin Recht. Gradient descent converges to minimizers. University of California, Berkeley, 1050:16, 2016.
Mu Li, Tong Zhang, Yuqiang Chen, and Alexander J Smola. Efficient mini-batch training for stochastic optimization. In Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining, pp. 661–670. ACM, 2014.
David JC MacKay. A practical bayesian framework for backpropagation networks. Neural compu- tation, 4(3):448–472, 1992.
Volodymyr Mnih, Koray Kavukcuoglu, David Silver, Alex Graves, Ioannis Antonoglou, Daan Wier- stra, and Martin Riedmiller. Playing atari with deep reinforcement learning. arXiv preprint arXiv:1312.5602, 2013.
Hossein Mobahi. Training recurrent neural networks by diffusion. arXiv preprint arXiv:1601.04114, 2016.
Daniel Povey, Arnab Ghoshal, Gilles Boulianne, Lukas Burget, Ondrej Glembek, Nagendra Goel, Mirko Hannemann, Petr Motlicek, Yanmin Qian, Petr Schwarz, et al. The kaldi speech recognition toolkit. In IEEE 2011 workshop on automatic speech recognition and understanding, number EPFL-CONF-192584. IEEE Signal Processing Society, 2011.
Jorma Rissanen. A universal prior for integers and estimation by minimum description length. The Annals of statistics, pp. 416–431, 1983.
Uri Shaham, Yutaro Yamada, and Sahand Negahban. Understanding adversarial training: Increasing local stability of neural nets through robust optimization. arXiv preprint arXiv:1511.05432, 2015.
Karen Simonyan and Andrew Zisserman. Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556, 2014.
Daniel Soudry and Yair Carmon. No bad local minima: Data independent training error guarantees for multilayer neural networks. arXiv preprint arXiv:1605.08361, 2016.
Nitish Srivastava, Geoffrey E Hinton, Alex Krizhevsky, Ilya Sutskever, and Ruslan Salakhutdinov. Dropout: a simple way to prevent neural networks from overfitting. Journal of Machine Learning Research, 15(1):1929–1958, 2014.
I. Sutskever, J. Martens, G. Dahl, and G. Hinton. On the importance of initialization and momentum in deep learning. In Proceedings of the 30th International Conference on Machine Learning (ICML 2013), pp. 1139–1147, 2013.
11

Published as a conference paper at ICLR 2017
Sixin Zhang, Anna E Choromanska, and Yann LeCun. Deep learning with elastic averaging sgd. In Advances in Neural Information Processing Systems, pp. 685–693, 2015.
Stephan Zheng, Yang Song, Thomas Leung, and Ian Goodfellow. Improving the robustness of deep neural networks via stability training. arXiv preprint arXiv:1604.04326, 2016.
A DETAILS ABOUT DATA SETS
We summarize the data sets used in our experiments in Table 5. TIMIT is a speech recognition data set which is pre-processed using Kaldi (Povey et al., 2011) and trained using a fully-connected network. The rest of the data sets are used without any pre-processing.
Table 5: Data Sets
# Data Points
Data Set MNIST TIMIT CIFAR-10 CIFAR-100
Train
60000
721329
50000
50000
Test
10000
310621
10000
10000
# Features
28 × 28 360 32 × 32 32 × 32
# Classes
10 1973 10 100
Reference
(LeCun et al., 1998a;b) (Garofolo et al., 1993) (Krizhevsky & Hinton, 2009) (Krizhevsky & Hinton, 2009)
B B.1
ARCHITECTURE OF NETWORKS NETWORK F1
For this network, we use a 784-dimensional input layer followed by 5 batch-normalized (Ioffe & Szegedy, 2015) layers of 512 neurons each with ReLU activations. The output layer consists of 10 neurons with the softmax activation.
B.2 NETWORK F2
The network architecture for F2 is similar to F1. We use a 360-dimensional input layer followed by 7 batch-normalized layers of 512 neurons with ReLU activation. The output layer consists of 1973 neurons with the softmax activation.
B.3 NETWORKS C1 AND C3
The C1 network is a modified version of the popular AlexNet configuration (Krizhevsky et al., 2012). For simplicity, denote a stack of n convolution layers of a filters and a Kernel size of b × c with stride length of d as n × [a, b, c, d]. The C1 configuration uses 2 sets of [64, 5, 5, 2]–MaxPool(3) followed by 2 dense layers of sizes (384, 192) and finally, an output layer of size 10. We use batch- normalization for all layers and ReLU activations. We also use Dropout (Srivastava et al., 2014) of 0.5 retention probability for the two dense layers. The configuration C3 is identical to C1 except it uses 100 softmax outputs instead of 10.
B.4 NETWORKS C2 AND C4
The C2 network is a modified version of the popular VGG configuration (Simonyan & Zisserman, 2014). The C3 network uses the configuration: 2×[64, 3, 3, 1], 2×[128, 3, 3, 1], 3×[256, 3, 3, 1], 3× [512, 3, 3, 1], 3 × [512, 3, 3, 1] which a MaxPool(2) after each stack. This stack is followed by a 512- dimensional dense layer and finally, a 10-dimensional output layer. The activation and properties of each layer is as in B.3. As is the case with C3 and C1, the configuration C4 is identical to C2 except that it uses 100 softmax outputs instead of 10.
12

Published as a conference paper at ICLR 2017
C PERFORMANCE MODEL
As mentioned in Section 1, a training algorithm that operates in the large-batch regime without suffering from a generalization gap would have the ability to scale to much larger number of nodes than is currently possible. Such and algorithm might also improve training time through faster convergence. We present an idealized performance model that demonstrates our goal.
For LB method to be competitive with SB method, the LB method must (i) converge to minimizers that generalize well, and (ii) do it in a reasonably number of iterations, which we analyze here. Let Is and Il be number of iterations required by SB and LB methods to reach the point of comparable test accuracy, respectively. Let Bs and Bl be corresponding batch sizes and P be number of processors being used for training. Assume that P < Bl, and let fs(P) be the parallel efficiency of the SB method. For simplicity, we assume that fl(P), the parallel efficiency of the LB method, is 1.0. In other words, we assume that the LB method is perfectly scalable due to use of a large batch size. For LB to be faster than SB, we must have IBl 0, these techniques solve the problem
min φ(x):= max f(x+∆x) (6) x ∥∆x∥≤ε
Geometrically, classical (nominal) optimization attempts to locate the lowest point of a valley, while robust optimization attempts to lower an ε–disc down the loss surface. We refer an interested reader to (Bertsimas et al., 2010), and the references therein, for a review of non-convex robust optimiza- tion. A direct application of this technique is, however, not feasible in our context since each itera- tion is prohibitively expensive because it involves solving a large-scale second-order conic program (SOCP).
15