Description: WeightWatcher (WW): is an open-source, diagnostic tool for analyzing Deep Neural Networks (DNN), without needing access to training or even test data. It can be used to:analyze pre/trained PyTorch, Keras, DNN models (Conv2D and Dense layers) monitor models, and the model layers, to see if they are over-trained or over-parameterized, predict test accuracies across different models, with or without training data, and detect potential problems when compressing or fine-tuning pre-trained models. see https://weightwatcher.ai

1. calculation | consulting
weightwatcher - a diagnostic tool
for deep neural networks
(TM)
c|c
(TM)
charles@calculationconsulting.com

2. calculation|consulting
weightwatcher - a diagnostic tool
for deep neural networks
(TM)
charles@calculationconsulting.com

3. calculation | consulting why deep learning works
Who Are We?
c|c
(TM)
Dr. Charles H. Martin, PhD
University of Chicago, Chemical Physics
NSF Fellow in Theoretical Chemistry, UIUC
Over 15 years experience in applied Machine Learning and AI
ML algos for: Aardvark, acquired by Google (2010)
Demand Media (eHow); first $1B IPO since Google
Wall Street: BlackRock
Fortune 500: Roche, France Telecom
BigTech: eBay, Aardvark (Google), GoDaddy
Private Equity: Griffin Advisors
Alt. Energy: Anthropocene Institute (Page Family)
www.calculationconsulting.com
charles@calculationconsulting.com
(TM)
3

4. calculation | consulting why deep learning works
c|c
(TM)
(TM)
4
Michael W. Mahoney
ICSI, RISELab, Dept. of Statistics UC Berkeley
Algorithmic and statistical aspects of modern large-scale data analysis.
large-scale machine learning | randomized linear algebra
geometric network analysis | scalable implicit regularization
PhD, Yale University, computational chemical physics
SAMSI National Advisory Committee
NRC Committee on the Analysis of Massive Data
Simons Institute Fall 2013 and 2018 program on the Foundations of Data
Biennial MMDS Workshops on Algorithms for Modern Massive Data Sets
NSF/TRIPODS-funded Foundations of Data Analysis Institute at UC Berkeley
https://www.stat.berkeley.edu/~mmahoney/
mmahoney@stat.berkeley.edu
Who Are We?

5. c|c
(TM)
(TM)
5
calculation | consulting why deep learning works
Understanding deep learning requires rethinking generalization
Motivations: WeightWatcher Theory
The weightwatcher theory is a Semi-Empirical theory based on:
the Statistical Mechanics of Generalization,
Random MatrixTheory, and
the theory of Strongly Correlated Systems
Great nerdy stuff, however, I will be discussing the weightwatcher tool
and what it can do for you

6. c|c
(TM)
(TM)
6
calculation | consulting why deep learning works
Open source tool: weightwatcher

7. c|c
(TM)
(TM)
7
calculation | consulting why deep learning works
Open source tool: weightwatcher
pip install weightwatcher
…
import weightwatcher as ww
watcher = ww.WeightWatcher(model=model)
results = watcher.analyze()
watcher.get_summary()
watcher.print_results()
https://github.com/CalculatedContent/WeightWatcher

8. c|c
(TM)
WeightWatcher: A diagnostic tool
(TM)
8
calculation | consulting why deep learning works
Analyze pre/trained pyTorch, TF/Keras, and ONNX models
Inspect models that are difficult to train
Gauge improvements in model performance
Predict test accuracies across different models
Detect problems when compressing or fine-tuning pretrained models
pip install weightwatcher

9. c|c
(TM)
Research: Implicit Self-Regularization in Deep Learning
(TM)
9
calculation | consulting why deep learning works
• Implicit Self-Regularization in Deep Neural Networks: Evidence from
Random Matrix Theory and Implications for Learning.
• Heavy-Tailed Universality Predicts Trends in Test Accuracies for Very Large
Pre-Trained Deep Neural Networks
• Workshop: Statistical Mechanics Methods for Discovering Knowledge from
Production-Scale Neural Networks
• Predicting trends in the quality of state-of-the-art neural networks without
access to training or testing data
• (more submitted and in progress today)
(JMLR 2021)
(ICML 2019)
(KDD 2020)
(Nature Communications 2021)
Selected publications

10. c|c
(TM)
WeightWatcher: analyzes the ESD
(eigenvalues) of the layer weight matrices
(TM)
10
calculation | consulting why deep learning works
The tail of the ESD contains the information

11. c|c
(TM)
(TM)
11
calculation | consulting why deep learning works
ESD of DNNs: detailed insight into W
Empirical Spectral Density (ESD: eigenvalues of X)
import keras
import numpy as np
import matplotlib.pyplot as plt
…
W = model.layers[i].get_weights()[0]
N,M = W.shape()
…
X = np.dot(W.T, W)/N
evals = np.linalg.eigvals(X)
plt.hist(evals, bin=100, density=True)

12. c|c
(TM)
WeightWatcher: analyzes the ESD
(eigenvalues) of the layer weight matrices
(TM)
12
calculation | consulting why deep learning works
details_df = watcher.analyze(model=your_model)
The tool provides various plots, quality metrics, and transforms

13. c|c
(TM)
WeightWatcher: analyzes the ESD
(eigenvalues) of the layer weight matrices
(TM)
13
calculation | consulting why deep learning works
watcher.analyze(…, plot=True, …)

14. c|c
(TM)
WeightWatcher: analyzes the ESD
(eigenvalues) of the layer weight matrices
(TM)
14
calculation | consulting why deep learning works
Well trained laters are heavy-tailed and well shaped
GPT-2 Fits a Power Law
(or Truncated Power Law)
alpha in [2, 6]
watcher.analyze(plot=True)
Good quality of fit (D is small)

15. c|c
(TM)
WeightWatcher: analyzes the ESD
(eigenvalues) of the layer weight matrices
(TM)
15
calculation | consulting why deep learning works
Better trained laters are more heavy-tailed and better shaped
GPT GPT-2

16. c|c
(TM)
(TM)
16
calculation | consulting why deep learning works
Heavy Tailed Metrics: GPT vs GPT2
The original GPT is poorly trained on purpose; GPT2 is well trained
alpha for every layer
smaller alpha is better
large alpha are bad fits

17. c|c
(TM)
(TM)
17
calculation | consulting why deep learning works
Power Law Universality: ImageNet
All ImageNet models display remarkable Heavy Tailed Universality
500 matrices
~50 architectures
Linear layers &
Conv2D feature maps
80-90% < 4

18. c|c
(TM)
(TM)
18
calculation | consulting why deep learning works
Random Matrix Theory: detailed insight into WL
DNN training induces breakdown of Gaussian random structure
and the onset of a new kind of heavy tailed self-regularization
Gaussian
random
matrix
Bulk+
Spikes
Heavy
Tailed
Small, older NNs
Large, modern DNNs
and/or
Small batch sizes

19. c|c
(TM)
(TM)
19
calculation | consulting why deep learning works
Random Matrix Theory: Marcenko Pastur
plus Tracy-Widom fluctuations
very crisp edges
Q
RMT says if W is a simple random Gaussian matrix,
then the ESD will have a very simple , known form
Shape depends on Q=N/M
(and variance ~ 1)
Eigenvalues tightly bounded
a few spikes may appear

20. c|c
(TM)
(TM)
20
calculation | consulting why deep learning works
RMT: AlexNet
Marchenko-Pastur Bulk-decay | Heavy Tailed
FC1
zoomed in
FC2
zoomed in

21. c|c
(TM)
(TM)
21
calculation | consulting why deep learning works
Random Matrix Theory: Heavy Tailed
But if W is heavy tailed, the ESD will also have heavy tails
(i.e. its all spikes, bulk vanishes)
If W is strongly correlated , then the ESD can be modeled as if W is drawn
from a heavy tailed distribution
Nearly all pre-trained DNNs display heavy tails…as shall soon see

22. c|c
(TM)
(TM)
22
calculation | consulting why deep learning works
AlexNet,
VGG,
ResNet,
Inception,
DenseNet,
…
Heavy Tailed RMT: Scale Free ESD
All large, well trained, modern DNNs exhibit heavy tailed self-regularization
scale free

23. c|c
(TM)
(TM)
23
calculation | consulting why deep learning works
HT-SR Theory: 5+1 Phases of Training
Implicit Self-Regularization in Deep Neural Networks: Evidence from Random Matrix Theory and Implications for Learning
Charles H. Martin, Michael W. Mahoney; JMLR 22(165):1−73, 2021.

24. c|c
(TM)
(TM)
24
calculation | consulting why deep learning works
Heavy Tailed RMT: Universality Classes
The familiar Wigner/MP Gaussian class is not the only Universality class in RMT

25. c|c
(TM)
WeightWatcher: predict trends in generalization
(TM)
25
calculation | consulting why deep learning works
Predict test accuracies across variations in hyper-parameters
The average Power Law exponent alpha
predicts generalization—at fixed depth
Smaller average-alpha is better
Better models are easier to treat

26. c|c
(TM)
WeightWatcher: Shape vs Scale metrics
(TM)
26
calculation | consulting why deep learning works
Purely norm-based (scale) metrics (from SLT) can be correlated with depth
but anti-correlated with hyper-parameter changes

27. c|c
(TM)
WeightWatcher: treat architecture changes
(TM)
27
calculation | consulting why deep learning works
Predict test accuracies across variations in hyper-parameters and depth
The alpha-hat metric combines
shape and scale metrics
and corrects
for different depths (grey line)
can be derived from theory…

28. c|c
(TM)
WeightWatcher: predict test accuracies
(TM)
28
calculation | consulting why deep learning works
alpha-hat works for 100s of different CV and NLP models
(Nature Communications 2021)
We do not have access to
The training or test data
But we can still predict
trends in the generalization

29. c|c
(TM)
(TM)
29
calculation | consulting why deep learning works
Predicting test accuracies: Heavy-tailed shape metrics
The heavy tailed metrics perform best
Weighted Alpha Alpha (Shatten) Norm

30. c|c
(TM)
WeightWatcher: predict test accuracies
(TM)
30
calculation | consulting why deep learning works
ResNet, DenseNet, etc.
(Nature Communications 2021)

31. c|c
(TM)
(TM)
31
calculation | consulting why deep learning works

32. c|c
(TM)
(TM)
32
calculation | consulting why deep learning works
Experiments: just apply to pre-trained Models
LeNet5 (1998)
AlexNet (2012)
InceptionV3 (2014)
ResNet (2015)
…
DenseNet201 (2018)
https://medium.com/@siddharthdas_32104/
cnns-architectures-lenet-alexnet-vgg-googlenet-resnet-and-more-666091488df5
Conv2D MaxPool Conv2D MaxPool FC FC

33. c|c
(TM)
(TM)
33
calculation | consulting why deep learning works
Predicting test accuracies: 100 pretrained models
The heavy tailed metrics perform best
https://github.com/osmr/imgclsmob
From an open source sandbox of
nearly 500 pretrained CV models
(picked >=5 models / regression)

34. c|c
(TM)
(TM)
34
calculation | consulting why deep learning works
Correlation Flow: CV Models
We can study correlation flow looking at vs. depth
VGG ResNet DenseNet

35. c|c
(TM)
WeightWatcher: detect overfitting
(TM)
35
calculation | consulting why deep learning works
Provide a data-independent criteria for early-stopping

36. c|c
(TM)
WeightWatcher: global and local convexity metrics
(TM)
36
calculation | consulting why deep learning works
Smaller alpha corresponds to more convex energy landscapes
Transformers (alpha ~ 3-4 or more)
alpha 2-3 (or less)
Rational Decisions, Random Matrices and Spin Glasses" (1998)
by Galluccio, Bouchaud, and Potters:

37. c|c
(TM)
WeightWatcher: global and local convexity metrics
(TM)
37
calculation | consulting why deep learning works
When the layer alpha < 2, we think this means the layer is overfit
We suspect that the early layers
of some Convolutional Nets
may be slightly overtrained
Some alpha < 2
This is predicted from our HTSR theory

38. c|c
(TM)
WeightWatcher: scale and shape anomalies
(TM)
38
calculation | consulting why deep learning works
We can detect problems in layers not detectable otherwise

39. c|c
(TM)
(TM)
39
calculation | consulting why deep learning works
Detect potential signatures of over-fitting
WeightWatcher: Correlation Traps
watcher.analyze(plot=True, randomize=True)

40. c|c
(TM)
WeightWatcher: SVDSharpness transform
(TM)
40
calculation | consulting why deep learning works
Remove potential signatures of over-fitting
Like PAC-bounds Sharpness transform
Clips bad elements in W using RMT
clip
smoothed_model =
watcher.SVDSharpness(model=your_model)

41. c|c
(TM)
WeightWatcher: RMT-based shape metrics
(TM)
41
calculation | consulting why deep learning works
ww also includes predictive, non-parametric shape metrics
rand_distance =
jensen_shannon_distance(original_esd, random_esd)

42. c|c
(TM)
WeightWatcher: RMT-based shape metrics
(TM)
42
calculation | consulting why deep learning works
the layer rand_distance and alpha metrics are correlated

43. c|c
(TM)
WeightWatcher: interpreting shapes
(TM)
43
calculation | consulting why deep learning works
very high accuracy requires advanced methods
hard
easy
X
:)

44. c|c
(TM)
WeightWatcher: more Power Law shape metrics
(TM)
44
calculation | consulting why deep learning works
watcher.analyze(…, fit=‘TPL’)
Truncated Power Law fits
fit=‘E_TPL’)
weightwatcher provides several shape (and scale) metrics
plus several more unpublished experimental options

45. c|c
(TM)
WeightWatcher: E_TPL shape metric
(TM)
45
calculation | consulting why deep learning works
the E_TPL (and rand_distance)
shape metrics track the
learning curve epoch-by-epoch
Training MT transformers
from scratch to SOTA
Extended Truncated Power Law
highly accurate results leverage the advanced shape metrics
Here, (Lambda) is the shape metric

46. c|c
(TM)
WeightWatcher: why Power Law fits ?
(TM)
46
calculation | consulting why deep learning works
Spiking (i.e real) neurons exhibit power law behavior
weightwatcher supports several PL fits
from experimental neuroscience
plus totally new shape metrics
we have invented (and published)

47. c|c
(TM)
WeightWatcher: why Power Law fits ?
(TM)
47
calculation | consulting why deep learning works
Spiking (i.e real) neurons exhibit (truncated) power law behavior
The Critical Brain Hypothesis
Evidence of Self-Organized Criticality (SOC)
Per Bak (How Nature Works)
As neural systems become more complex
they exhibit power law behavior
and then truncated power law behavior
We see exactly this behavior in DNNs
and it is predictive of learning capacity

48. c|c
(TM)
WeightWatcher: open-source, open-science
(TM)
48
calculation | consulting why deep learning works
We are looking for early adopters and collaborators
github.com/CalculatedContent/WeightWatcher
We have a Slack channel to support the tool
Please file issues
Ping me to join

49. c|c
(TM)
(TM)
49
calculation | consulting why deep learning works
Statistical Mechanics
derivation of the alpha-hat metric

50. c|c
(TM)
(TM)
50
calculation | consulting why deep learning works
Classic Set Up: Student-Teacher model
Statistical Mechanics of Learning Engle &Van den Broeck (2001)
Generalization error ~ phase space volume
Average error ~ overlap between T and J

51. c|c
(TM)
(TM)
51
calculation | consulting why deep learning works
Classic Set Up: Student-Teacher model
Statistical Mechanics of Learning Engle &Van den Broeck (2001)
Standard approach:
• Teacher (T) and Student (J) random Perceptron vectors
• Treat data as an external random Gaussian field
• Apply Hubbard–Stratonovich to get mean-field result
• Assume continuous or discrete J
• Solve for as a function of load (# data points / # parameters)

52. c|c
(TM)
(TM)
52
calculation | consulting why deep learning works
New Set Up: Matrix-generalized Student-Teacher
x
Continuous perceptron
Ising Perceptron
Uninteresting
Replica theory, shows phase behavior,
entropy collapse, etc

53. c|c
(TM)
(TM)
53
calculation | consulting why deep learning works
New Set Up: Matrix-generalized Student-Teacher
“Towards a new theory…” Martin, Milletari, & Mahoney (in preparation)
real DNN matrices:
NxM
Strongly correlated
Heavy tailed
correlation matrices
Solve for total integrated phase-space volume

54. c|c
(TM)
(TM)
54
calculation | consulting why deep learning works
New approach: HCIZ Matrix Integrals
Write the overlap as
Fix a Teacher. The integral is now over all random students J that overlap w/T
Use the following result in RMT
“Asymptotics of HCZI integrals …” Tanaka (2008)

55. c|c
(TM)
(TM)
55
calculation | consulting why deep learning works
RMT: Annealed vs Quenched averages
“A First Course in Random Matrix Theory” Potters and Bouchaud (2020)
good outside spin-glass phases
where system is trained well
We imagine averaging over all (random) students DNNs
with (correlations that) look like the teacher DNN

56. c|c
(TM)
(TM)
56
calculation | consulting why deep learning works
New interpretation: HCIZ Matrix Integrals
Generating functional
R-Transform (inverse Green’s function, via Contour Integral)
in terms of Teacher's eigenvalues , and Student’s cumulants

57. c|c
(TM)
(TM)
57
calculation | consulting why deep learning works
Some basic RMT: Greens functions
The Green’s function is the Stieltjes transform of eigenvalue distribution
Given the empirical spectral density (average eigenvalue density)
and using:

58. c|c
(TM)
(TM)
58
calculation | consulting why deep learning works
Some basic RMT: Moment generating functions
The Green’s function has poles
at the actual eigenvalues
But is analytic in the complex plane
up and away from the real axis
z
Expand in a series around z = ∞
moment generating function

59. c|c
(TM)
(TM)
59
calculation | consulting why deep learning works
Some basic RMT: R-Transforms
Which gives a (similar) moment generating function
The free-cumulant-generating function (R-transform)
is related to the Green function as
Gaussian random and very Heavy Tailed (Levy) random matrices
but which takes a simple form for both

60. c|c
(TM)
(TM)
60
calculation | consulting why deep learning works
Results: Gaussian Random Weight Matrices
“Random Matrix Theory (book)” Bouchaud and Potters (2020)
Recover the Frobenius Norm (squared) as the metric

61. c|c
(TM)
(TM)
61
calculation | consulting why deep learning works
Results: (very) Heavy Tailed Weight Matrices
“Heavy-tailed random matrices” Burda and Jukiewicz (2009)
Recover a Shatten Norm, in terms of the Heavy Tailed exponent

62. c|c
(TM)
(TM)
62
calculation | consulting why deep learning works
Application to: Heavy Tailed Weight Matrices
Some reasonable approximations give the weighted alpha metric
Q.E.D.

63. (TM)
c|c
(TM)
c | c
charles@calculationconsulting.com