-
A Scalable, Interpretable, Verifiable & Differentiable Logic Gate Convolutional Neural Network Architecture From Truth Tables
Authors:
Adrien Benamira,
Tristan Guérand,
Thomas Peyrin,
Trevor Yap,
Bryan Hooi
Abstract:
We propose $\mathcal{T}$ruth $\mathcal{T}$able net ($\mathcal{TT}$net), a novel Convolutional Neural Network (CNN) architecture that addresses, by design, the open challenges of interpretability, formal verification, and logic gate conversion. $\mathcal{TT}$net is built using CNNs' filters that are equivalent to tractable truth tables and that we call Learning Truth Table (LTT) blocks. The dual fo…
▽ More
We propose $\mathcal{T}$ruth $\mathcal{T}$able net ($\mathcal{TT}$net), a novel Convolutional Neural Network (CNN) architecture that addresses, by design, the open challenges of interpretability, formal verification, and logic gate conversion. $\mathcal{TT}$net is built using CNNs' filters that are equivalent to tractable truth tables and that we call Learning Truth Table (LTT) blocks. The dual form of LTT blocks allows the truth tables to be easily trained with gradient descent and makes these CNNs easy to interpret, verify and infer. Specifically, $\mathcal{TT}$net is a deep CNN model that can be automatically represented, after post-training transformation, as a sum of Boolean decision trees, or as a sum of Disjunctive/Conjunctive Normal Form (DNF/CNF) formulas, or as a compact Boolean logic circuit. We demonstrate the effectiveness and scalability of $\mathcal{TT}$net on multiple datasets, showing comparable interpretability to decision trees, fast complete/sound formal verification, and scalable logic gate representation, all compared to state-of-the-art methods. We believe this work represents a step towards making CNNs more transparent and trustworthy for real-world critical applications.
△ Less
Submitted 2 February, 2023; v1 submitted 17 August, 2022;
originally announced August 2022.
-
The Macroporosity of Rubble Pile Asteroid Ryugu and Implications for the Origin of Chondrules
Authors:
William Herbst,
James P. Greenwood,
Teng Ee Yap
Abstract:
We use the known surface boulder-size distribution of the C-type rubble pile asteroid Ryugu (NEA 162173) to determine its macroporosity, assuming it is a homogeneous granular aggregate. We show that the volume-frequency distribution of its boulders, cobbles and pebbles, is well represented by a lognormal function with $σ= 2.4 \pm 0.1$ and $μ= 0.2 \pm 0.05$. Application of linear-mixture packing th…
▽ More
We use the known surface boulder-size distribution of the C-type rubble pile asteroid Ryugu (NEA 162173) to determine its macroporosity, assuming it is a homogeneous granular aggregate. We show that the volume-frequency distribution of its boulders, cobbles and pebbles, is well represented by a lognormal function with $σ= 2.4 \pm 0.1$ and $μ= 0.2 \pm 0.05$. Application of linear-mixture packing theory yields a value for the macroporosity of $φ= 0.14 \pm 0.04$. Given its low bulk density of 1.19 gm cm$^{-3}$, this implies an average density for Ryugu's rocks of $1.38 \pm 0.07$ gm cm$^{-3}$ throughout its volume, consistent with a recent determination for surface boulders based on their thermal properties. This supports the spectrum-based argument that IDP's may be the best analog material available on Earth and suggests that high-density, well-lithified objects such as chondrules and chondrule-bearing chondrites may be rare on Ryugu. Implications of this result for the origin of chondrules, a long-standing problem in cosmochemistry, are discussed. We propose that chondrules and most chondrites formed together in rare lithification events, which occurred during the accretion of chondritic envelopes to large, differentiated planetesimals at a time when they were still hot from $^{26}$Al decay.
△ Less
Submitted 13 April, 2021;
originally announced April 2021.
-
Permutations Avoiding Certain Partially-ordered Patterns
Authors:
Kai Ting Keshia Yap,
David Wehlau,
Imed Zaguia
Abstract:
A permutation $π$ contains a pattern $σ$ if and only if there is a subsequence in $π$ with its letters are in the same relative order as those in $σ$. Partially ordered patterns (POPs) provide a convenient way to denote patterns in which the relative order of some of the letters does not matter. This paper elucidates connections between the avoidance sets of a few POPs with other combinatorial obj…
▽ More
A permutation $π$ contains a pattern $σ$ if and only if there is a subsequence in $π$ with its letters are in the same relative order as those in $σ$. Partially ordered patterns (POPs) provide a convenient way to denote patterns in which the relative order of some of the letters does not matter. This paper elucidates connections between the avoidance sets of a few POPs with other combinatorial objects, directly answering five open questions posed by Gao and Kitaev \cite{gao-kitaev-2019}. This was done by thoroughly analysing the avoidance sets and develo** recursive algorithms to derive these sets and their corresponding combinatorial objects in parallel, which yielded a natural bijection. We also analysed an avoidance set whose simple permutations are enumerated by the Fibonacci numbers and derived an algorithm to obtain them recursively.
△ Less
Submitted 28 January, 2021;
originally announced January 2021.
-
Full Automation for Rapid Modulator Characterization and Accurate Analysis Using SciPy
Authors:
T. L. Yap,
A. Sasidhara,
N. X. Ang,
X. Guo,
W. Wang,
K. S. Ang,
S. L. Tan
Abstract:
Modulator testing involved complex biasing conditions, hardware connections and data analysis. Also, any optical signal distortion due to the grating coupler effect could potentially induce additional difficulty in setting the correct bias condition for an accurate measurement of the modulator performance. In this paper, we proposed to use SciPy, an open-source scientific computing library, for au…
▽ More
Modulator testing involved complex biasing conditions, hardware connections and data analysis. Also, any optical signal distortion due to the grating coupler effect could potentially induce additional difficulty in setting the correct bias condition for an accurate measurement of the modulator performance. In this paper, we proposed to use SciPy, an open-source scientific computing library, for automation in the silicon modulator test with bias setting and data analysis.
△ Less
Submitted 11 October, 2020;
originally announced October 2020.
-
Training CNN Classifiers for Semantic Segmentation using Partially Annotated Images: with Application on Human Thigh and Calf MRI
Authors:
Chun Kit Wong,
Stephanie Marchesseau,
Maria Kalimeri,
Tiang Siew Yap,
Serena S. H. Teo,
Lingaraj Krishna,
Alfredo Franco-Obregón,
Stacey K. H. Tay,
Chin Meng Khoo,
Philip T. H. Lee,
Melvin K. S. Leow,
John J. Totman,
Mary C. Stephenson
Abstract:
Objective: Medical image datasets with pixel-level labels tend to have a limited number of organ or tissue label classes annotated, even when the images have wide anatomical coverage. With supervised learning, multiple classifiers are usually needed given these partially annotated datasets. In this work, we propose a set of strategies to train one single classifier in segmenting all label classes…
▽ More
Objective: Medical image datasets with pixel-level labels tend to have a limited number of organ or tissue label classes annotated, even when the images have wide anatomical coverage. With supervised learning, multiple classifiers are usually needed given these partially annotated datasets. In this work, we propose a set of strategies to train one single classifier in segmenting all label classes that are heterogeneously annotated across multiple datasets without moving into semi-supervised learning. Methods: Masks were first created from each label image through a process we termed presence masking. Three presence masking modes were evaluated, differing mainly in weightage assigned to the annotated and unannotated classes. These masks were then applied to the loss function during training to remove the influence of unannotated classes. Results: Evaluation against publicly available CT datasets shows that presence masking is a viable method for training class-generic classifiers. Our class-generic classifier can perform as well as multiple class-specific classifiers combined, while the training duration is similar to that required for one class-specific classifier. Furthermore, the class-generic classifier can outperform the class-specific classifiers when trained on smaller datasets. Finally, consistent results are observed from evaluations against human thigh and calf MRI datasets collected in-house. Conclusion: The evaluation outcomes show that presence masking is capable of significantly improving both training and inference efficiency across imaging modalities and anatomical regions. Improved performance may even be observed on small datasets. Significance: Presence masking strategies can reduce the computational resources and costs involved in manual medical image annotations. All codes are publicly available at https://github.com/wong-ck/DeepSegment.
△ Less
Submitted 16 August, 2020;
originally announced August 2020.
-
Low Temperature Nanoscale Electronic Transport on the MoS_2 surface
Authors:
R. Thamankar,
T. L. Yap,
K. E. J. Goh,
C. Troadec,
C. Joachim
Abstract:
Two-probe electronic transport measurements on a Molybdenum Disulphide (MoS_2) surface were performed at low temperature (30K) under ultra-high vacuum conditions. Two scanning tunneling microscope tips were precisely positioned in tunneling contact to measure the surface current-voltage characteristics. The separation between the tips is controllably varied and measured using a high resolution sca…
▽ More
Two-probe electronic transport measurements on a Molybdenum Disulphide (MoS_2) surface were performed at low temperature (30K) under ultra-high vacuum conditions. Two scanning tunneling microscope tips were precisely positioned in tunneling contact to measure the surface current-voltage characteristics. The separation between the tips is controllably varied and measured using a high resolution scanning electron microscope. The MoS_2 surface shows a surface electronic gap (E_S) of 1.4eV measured at a probe separation of 50nm. Furthermore, the two- probe resistance measured outside the electronic gap shows 2D-like behavior with the two-probe separation.
△ Less
Submitted 14 August, 2013;
originally announced August 2013.
