arXiv subjects

Haichen Zhou

Publications and source records attributed to Haichen Zhou.

6 recordsLinked to original sources

arXiv2026

Alleviating Regional Shortcuts for Few-Shot Class-Incremental Learning

Few-shot class-incremental learning (FSCIL) aims to incrementally learn novel classes with only a few samples while avoiding forgetting base classes. However, current methods show a tendency to misclassify novel-class samples into base classes, which we find to be caused by the excessive focus on base-class-discriminative regions on novel-class samples. In this work, we aim to explore the underlying mechanism for an interpretation and solution. We first provide a compositional view to analyze the transferred and reused spatial patterns on novel-class samples. Then, through extensive experiments and theoretical analysis, we identify both empirically and theoretically that a shortcut exists in the model's base-class training, which naturally forms the excessive focus on only the most discriminative regions (primitives), which we term as the regional shortcut. Finally, based on this interpretation, to address this problem, we propose a compositional-learning-based method to learn two primitive sets (a common set and a discriminative set), which alleviates the regional shortcut by constraining the model to learn and utilize the common primitive set for base- and novel-class recognition. Extensive experiments on standard FSCIL benchmarks demonstrate the effectiveness of our approach, yielding consistent improvements over existing state-of-the-art methods in both accuracy and interpretability.

cs.CV↗

arXiv2025

Open-Path Methane Sensing via Backscattered Light in a Nonlinear Interferometer

Nonlinear interferometry has widespread applications in sensing, spectroscopy, and imaging. However, most implementations require highly reflective mirrors and precise optical alignment, drastically reducing their versatility and usability in outdoor applications. This work is based on stimulated parametric down conversion (ST-PDC), demonstrating methane absorption spectroscopy in the mid-infrared (MIR) region by detecting near-infrared (NIR) photons using a silicon-based CMOS camera. The MIR light, used to probe methane, is diffusely backscattered from a Lambertian surface, experiencing significant transmission loss. We implement a single-mode confocal illumination and collection scheme, using a two-lens system to mode-match the interfering beams to achieve background methane detection at a distance of 4.6 meters under a 60 dB loss. Our method is also extended to real-world surfaces, such as glass, brushed metal, and a leaf, showing robust background methane sensing with various target materials.

physics.optics↗

arXiv2024

Methane Sensing via Unbalanced Nonlinear Interferometry using a CMOS Camera

Here we present a high-sensitivity, rapid, and low-cost method for methane sensing based on a nonlinear interferometer. This method utilizes signal photons generated by stimulated parametric down-conversion (ST-PDC), enabling the use of a silicon detector to capture high-precision methane absorption spectra in the mid-infrared region. By controlling the system loss, we achieve more significant changes in visibility, thereby increasing sensitivity. The methane concentration within a gas cell is determined accurately. In addition, ST-PDC enables long-distance sensing and the capability to measure low ambient methane concentrations in the real world. A low-cost CMOS camera is employed to capture spatial interference fringes, ensuring fast and efficient detection.

physics.optics↗

arXiv2024

Compositional Few-Shot Class-Incremental Learning

Few-shot class-incremental learning (FSCIL) is proposed to continually learn from novel classes with only a few samples after the (pre-)training on base classes with sufficient data. However, this remains a challenge. In contrast, humans can easily recognize novel classes with a few samples. Cognitive science demonstrates that an important component of such human capability is compositional learning. This involves identifying visual primitives from learned knowledge and then composing new concepts using these transferred primitives, making incremental learning both effective and interpretable. To imitate human compositional learning, we propose a cognitive-inspired method for the FSCIL task. We define and build a compositional model based on set similarities, and then equip it with a primitive composition module and a primitive reuse module. In the primitive composition module, we propose to utilize the Centered Kernel Alignment (CKA) similarity to approximate the similarity between primitive sets, allowing the training and evaluation based on primitive compositions. In the primitive reuse module, we enhance primitive reusability by classifying inputs based on primitives replaced with the closest primitives from other classes. Experiments on three datasets validate our method, showing it outperforms current state-of-the-art methods with improved interpretability. Our code is available at https://github.com/Zoilsen/Comp-FSCIL.

cs.CV↗

arXiv2024

Methane sensing in the mid-IR using short wave IR photon counting detectors via non-linear interferometry

We demonstrate a novel MIR methane sensor shifting measurement wavelength to SWIR (1.55$μ$m) by using non-linear interferometry. The technique exploits the interference effects seen in three-wave mixing when pump, signal, and idler modes make a double pass through a nonlinear crystal. The method allows sensing at wavelengths where detectors are poor ($>$3$μ$m) and detection at wavelengths where photon counting sensitivity can be achieved. In a first experimental demonstration, we measured a small methane concentration inside a gas cell with high precision. This interferometer can be built in a compact design for field operations and potentially enable the detection of low concentrations of methane at up to 100m range. Signal-to-noise ratio calculations show that the method can outperform existing short wavelength ($\sim$1.65$μ$m) integrated path differential absorption direct sensing at high ($>$$10^{-4}$) non-linear gain.

quant-ph↗

arXiv2024

Delve into Base-Novel Confusion: Redundancy Exploration for Few-Shot Class-Incremental Learning

Few-shot class-incremental learning (FSCIL) aims to acquire knowledge from novel classes with limited samples while retaining information about base classes. Existing methods address catastrophic forgetting and overfitting by freezing the feature extractor during novel-class learning. However, these methods usually tend to cause the confusion between base and novel classes, i.e., classifying novel-class samples into base classes. In this paper, we delve into this phenomenon to study its cause and solution. We first interpret the confusion as the collision between the novel-class and the base-class region in the feature space. Then, we find the collision is caused by the label-irrelevant redundancies within the base-class feature and pixel space. Through qualitative and quantitative experiments, we identify this redundancy as the shortcut in the base-class training, which can be decoupled to alleviate the collision. Based on this analysis, to alleviate the collision between base and novel classes, we propose a method for FSCIL named Redundancy Decoupling and Integration (RDI). RDI first decouples redundancies from base-class space to shrink the intra-base-class feature space. Then, it integrates the redundancies as a dummy class to enlarge the inter-base-class feature space. This process effectively compresses the base-class feature space, creating buffer space for novel classes and alleviating the model's confusion between the base and novel classes. Extensive experiments across benchmark datasets, including CIFAR-100, miniImageNet, and CUB-200-2011 demonstrate that our method achieves state-of-the-art performance.

cs.CV↗